<figure>
<img src="https://content.communere.com/assets/e2132209-57a8-47ea-97f8-f87e5e4ef36a" />
 <figcaption style="text-align: center;">
</figcaption>
</figure>

## Intro
Almost all entrepreneurs have experienced this scenario:
"Spending so much money and time on developing an app for your business only to find out no one is willing to download it!"
 But don’t get me wrong. It doesn’t mean your business app is a failure, but simply because nowadays, people are more comfortable using web applications.
 
The reasons are obvious: 
- These apps don’t use your device’s memory, RAM, CPU, etc.
- You don’t have to go through all the steps of downloading, installing, and updating them every now and then.
- They integrate with any device and any operating system.
- And more importantly, web app development costs far less than developing an app for your business. 

But there’s only a slight problem here for entrepreneurs: How to find the best web app developers to order your app?
In the following, we’ll outline all you need to know about the necessary features web application agencies must offer. Then we’ll go over how you can find the most professional developers for your business. Keep reading!

## What are the key features of web applications?
Of course, each application has its unique features based on the type of business and its goals. However, there are some must-have features that most professional web-based application developers offer. 

### Compatibility
One of the biggest mistakes most companies make is that they care more about displaying their application or website on a laptop than on a mobile phone. But today, most users prefer smartphones over laptops to access websites and applications. So, the most essential feature of a modern web application (WA) is to be compatible with mobile devices, tablets, and other devices besides laptops.

### Notifications
If you don’t want your users to miss out on the latest news, offers, updates, and so on, you’ll have to add push notifications to the list of necessary features. Push notifications make sure your users will get the information firsthand while they are doing their choirs. 
Although notifications can inform customers and increase your site traffic, they can have an undesirable effect if not used wisely. Too many notifications will only disturb your users. So, only use this feature for special occasions when you have something your audience will most likely appreciate.

### Strong security
Although security is the number three feature on this list, it must be your number one priority. The best web app developers, such as the Communere Ltd team, use the HTTPS protocol to secure users’ connections and transactions. In addition, HTTPS protocol ensures that your users’ accounts, identities, and communications are private and safe. And the safer your web app is, the more credit it gets from Google. So HTTPS can positively affect your SEO as well. 

### Add to the home screen
Add to the home screen is a unique feature in web apps that make them seem like native apps. With just two simple clicks, users can add any WA to their home screen and use them as a native app. However, they come with many advantages:
- Users don’t need to download, install, or update WPs, unlike native apps.
- This feature will increase users’ satisfaction as it helps them access the website with just a click. 
- Businesses can develop a web application and have a native application as well.
- It will improve users’ engagement rate.

### Search bar
If you own a small business, this feature might seem unnecessary. However, over time your business will grow, and you’ll end up with too much information on your WA, which gets hard to sort out daily. So, with all that, how do you expect your users to find anything on that web app?
That’s why a smart search bar is an important feature to add when developing a web-based application. For a better user experience, you can also add a filter section so your customers can search for their products or service more precisely.

### Integrated analytics
What is the most significant part of any business? Yes, it's data!
Why? Because it’s the most critical factor that can guarantee a business's success. With data, you can plan your next step and change your strategies if needed. It’s the only way to ensure your business is on the right path. 
As with any part of your business, you need to gather data and insights from your web app. That’s where integrated analytics comes to help. Instead of analysing each web app section, you can use its analytics to check its performance and how customers have responded. Then you can decide what changes are necessary to improve sales and take your business to the next level. 

### Touch or face ID
We all know how it feels like when you can’t remember your password to log into your account. You must go through many steps just to recover and change your password. But this problem was solved a long time ago. Now you can log in to any account in a matter of seconds. 
How? By touch or face ID. Unless you need a web application for businesses such as e-wallets, which require higher levels of security, integrated IDs can be a significant advantage.

![web app developers at Communere Ltd](https://content.communere.com/assets/5f80b1cc-d7cb-4ab5-95cd-fa00144a85c4)

## How to find the best enterprise web and app developers?
To find the best developers, you need to pay close attention to what they offer. In addition to the above features, finding a professional team that understands your requirements is vital. Based on customer reviews, the Communere Ltd team develop each web app based on businesses' goals and visions. Communere developers first try to understand the nature of your business and what you aim to achieve. Then, they’ll develop and design your web app based on your mission.
Moreover, a professional development company have:
- The latest technologies and tools
- Enough knowledge about your industry
- Reasonable prices for their services
- Excellent portfolios that you can find all over the internet
All of which you can find on the [Communere Ltd. Website](https://communere.com/). 

## Bottom line
Web app development is not something to be taken lightly. It will affect all aspects of your business. An enterprise web application directly impacts customers’ experience, satisfaction, loyalty, and, more importantly, how they see you within your industry.
Communere developers make sure all your requirements have been met. Communere's professional team first examines each business separately based on the type of industry, target audience, and target market. Then they design an exclusive web-based application so that it can have the most professional performance in addition to having a simple interface.

Features of best web app agencies

<figure>
<img src="https://content.communere.com/assets/086bf286-2119-4219-a6db-5854df0de33d" />
 <figcaption style="text-align: center;">
</figcaption>
</figure>

## Intro
One of the best and most popular ways to grow your business is by building a desktop, web, or mobile application to attract customers in the online world. Among the others, web applications are the favourite apps that help you achieve the best possible results.
This article will outline some of the critical benefits of web-based applications for businesses and provide information on how you can benefit from developing your own web application.
Increasing internet use in virtually every workplace has led to web applications becoming popular business tools. Their primary functions are communication with customers, collaboration with employees, secure data storage, and providing management with information and data.
A web app application is front-end-based, using the Internet as its interface, or 'front-end', to interact with users. The application can be accessed from any computer connected to the Internet or Intranet, and there is no need to install it on a stand-alone computer.
Today in this post, we are going to talk about some of the essential benefits of web applications. These advantages are all factors you need a business application to provide your customers with.

[Bespoke web app development](https://content.communere.com/assets/2c896f90-1887-41c3-9ec7-7ac103916cad)

## Benefits of a Web Application for Business
You and your business have the opportunity to benefit from a variety of advantages when you compare web-based applications to desktop applications. As these applications are available through the Internet, they can be accessed from any computer without having to manually install them on each computer.
Below we have mentioned some of the most critical web-based system advantages which will encourage you to think of web app development.

### The wide accessibility
One of the most important web-based operating systems advantages is accessibility across devices for users.
SaaS applications work consistently across all operating systems, regardless of whether the operating system has recently been updated or has been updated at the current time. This makes them useful for problem-solving because everyone knows how the application works.
Some good news, however, includes that most desktop applications can be converted into web applications to be accessed from anywhere and on any device. A flexible work schedule and location increase employee productivity within businesses when users are given the ability to choose when and where they want to use the application. Because of their web-based nature, business systems can be built using them.
All desktop-based applications must be updated on all of the devices on which they are installed to function correctly. In contrast, a company's web-based application can be updated in real-time without causing any downtime, enabling users to take advantage of the new features immediately.

### Easily customisable & scalable
There is often a limitation to off-the-shelf software when it comes to its ability to grow and integrate with clients' business operations. Many clients are unable to obtain these goals without the expenditure of additional expenditures. In addition, bespoke web applications can be tailored to meet the requirements of your business, making them wholly scalable and flexible.
You may also customise the application by rebranding it and assigning different permission levels to other users as part of the customisation process.
Because you are only using features and functions relevant to your business, web applications allow you to expand your application's functionality with your company's growth. Using a Web application that offers web-based functionality on mobile devices is possible if it is customised for their use.
In addition to being very user-friendly, the application can be built to the individual's preferences to present information in a personalised manner.
Web applications are much easier to be modified than desktop applications, which is a relatively rare occurrence. The advantage of web-based software is that employees can work with it on the go and access it whenever they need to since it can also be updated in response to changing needs.

[Customisable web applications](https://content.communere.com/assets/22e6b7bc-48ca-41b4-a10a-8cf74cddb125)

### It is fully integrated with other systems.
The ease of integrating web-based applications and other systems is far greater than desktop applications. The development of localised software is disadvantageous when compared to web-based applications that have a greater degree of interoperability. Thus, web-based applications are more readily interconnected than two completely independent systems running on different platforms.
Suppose you are still asking yourself why web-based applications should have such immense popularity. In that case, you must know that integration, full customisation, and wide accessibility are enough to encourage business owners to consider web apps.
If you agree that just these crucial pros of web apps are all a business need and if you are looking for expert web app developers, we recommend Communere Ltd. The Communere Ltd team builds robust, scalable, and bespoke web apps using agile development processes leveraging the latest web technologies.
Communere's developers are always learning new technologies and do everything they can to provide the means for them to update their knowledge.
You can also find more about our web app development service here:

 [Communere's web app service](https://communere.com/web-dev)
 
### Highly deployable
SaaS applications are accessible to businesses once they have been installed on the server hosting them. The host server can apply upgrades instead of updating each machine whenever an upgrade occurs.
The advantages of this approach include ease of implementation and efficient maintenance. Implementing a central point reduces the time to make changes so that the system remains consistent.
In light of the fact that web applications can now be managed and cross-platform supported, deploying and managing web applications within various work environments is now easier than ever before.
This service is the best choice when bandwidth is limited or the user uses a remote data server or system. You need only to send the user a URL where they may log in to the system. This is significant because it facilitates unrestricted access to a variety of systems, streamlines processes, and improves relationships between suppliers and customers.

### Increased flexibility
Upgrading your application is easy, and you can improve its flexibility if you wish to increase its growth capabilities. If you have several business processes that need to be completed simultaneously in your organisation, web-based software can definitely help you manage them more efficiently.
In addition, if one of your servers develops a problem, it can be replaced entirely without affecting your entire operating system, which will minimise the downtime you experience.

### Highly secure
Due to the security aspects of web-based applications, professionals and teams cannot access centra centralised without accessing the servers directly.
Cloud computing can make servers redundant and replicable to prevent downtime during a disaster. This approach eliminates the need for security on each device on which the application is installed, thereby simplifying the process.
As a result, there is a reduced likelihood of unauthorised access, and putting security measures in place is simpler since it can be done centrally. When desktop computers, particularly those running desktop software, are lost or damaged, replacing them can be very costly and time-consuming.
The critical thing to remember is that your data is at risk and that your software provider must reinstall the software on the new computer. You can be assured that if your computer equipment is damaged or stolen, you can quickly resume normal business operations with a web-based application stored in the cloud.
Using your URL username and password, you can securely access web-based applications from any computer or mobile device that has access to a remote server. The cloud can be used as a last resort to back up data in case of a loss of data due to any cause, regardless of whether it was caused by a human or by a computer program.

### Cost-effective
An organisation can substantially reduce operating costs by utilising web-based applications by reducing support and maintenance costs, reducing the load on end users' machines, and simplifying the application's architecture. A web-based application can save your company additional money since they streamline business processes.
The primary reason for developing a web application is that it takes less effort and time than developing a mobile application. Developing a web app also allows the application to be updated for multiple operating systems in one release. So It is a cost-effective method of implementing software since the application does not need to be tested on every operating system.

## Conclusion
Because web-based applications are easy to set up and manage, they can be a valuable tool for providing flexibility to your employees' work schedules, as well as having a number of other benefits.
The above are just several benefits that can be obtained from a web-based application and how they can assist businesses in achieving their objectives. When it comes to standing out from both online and offline competition, there is no doubt that you should consider developing a web application.
The Communere Ltd team is our recommendation for you to achieve success in the online world. This team uses modern management techniques to ensure your product is delivered on time. If you have any questions regarding web apps, you can quickly contact us by filling out this form:

[Contact Us Form](https://communere.com/contact)

Why Web Apps Are Necessary For Your Business? [ Key Benefits ]

<figure>
<img src="https://content.communere.com/assets/4e5c32ea-2c0e-4852-b328-4a01654dae8a" />
 <figcaption style="text-align: center;">
</figcaption>
</figure>

The way businesses are managed changed as a result of increased Internet usage by both corporations and individuals. Companies are now looking for professional web app developers for their web app development needs. But what exactly is a web app, and what are its features? 
In this article, we are going to provide you with a complete guide on web apps and explain why your business needs one. Keep reading!

## What's a web app?

A web app is a computer software that executes a specific task using a web browser. Some web apps need server-side processing because they are dynamic. Others require no server processing because they are entirely static. 
 
## Web applications vs other types of apps

Although they are frequently used interchangeably, websites and web applications are different. A website is most commonly described as a collection of interconnected, information-carrying pages that may be visited via a web browser. The user accesses a web application through a web browser or application software running on a web server.
Web apps are also different from native apps. Native applications are downloaded and built expressly for the sort of device they are installed on and utilise the hardware specific to the device, such as the GPS or camera.
Hybrid applications are occasionally used to describe programs that blend these two apps. Although hybrid apps are installed on the device similarly to native apps, they function similarly to web apps.

If you have a web app development project, Communere Ltd. makes sure you get the best experience. 

[Let's talk about your web app! ](https://communere.com/contact)

[https://content.communere.com/assets/3b7857c2-5cb5-496c-99ac-ab804a4df8ea]

## How does a web app work?

Web applications are written in two different languages by web app developers. To function, they often combine server-side scripts and client-side scripts. The information is stored and retrieved by the server-side script, which needs programming languages such as Python or Java. Languages like JavaScript, CSS, and HTML5 are required for the client-side script. These programming languages use the browser to carry out the operation.
The web server forwards the user's request to the web application server via the Internet when the user submits it through the application's user interface. The web app server completes the specified operation and generates the necessary data results. Then, these results are returned to the web server by the web app server.
Lastly, the user receives the required information from the web server on his mobile device, tablet, or desktop. 

## Most common features of web applications

The most common features of web apps are:
•	Prioritising user interaction
•	Easier to maintain compared to desktop apps
•	Operating online and through web browsers
•	Highly responsive
•	Using push notifications to keep users updated on app-related activity and encourage frequent app usage
•	No need to download or install them


## Most known web apps

Web app developers create different types of web apps, including email, word processors, spreadsheets, shopping carts, and more. Let's look at some of the most known ones available.

### Google Apps

Gmail, Google Docs, Google Slides, Google Sheets, online storage, and more are all included in Google Apps for Work. Document sharing is an essential feature of this well-known web application which enables simultaneous access to the same document version by all team members.

### Netflix

Netflix offers limitless on-demand access to movies and television series for its subscribers. It is also compatible with various devices and operating systems, like other online apps. Netflix serves as an example of why web apps are excellent content platforms.

### Microsoft Office

Microsoft Office has become a trendy online suite of tools, much like Google Workspace. You can also download mobile and desktop apps of Microsoft Office in addition to the web-based version.

### Basecamp

This well-known platform is a web application for project management, client messaging, and work schedules. It is simple to invite users to projects because it is a web app, so there is no need to worry about what kind of device or browser they are using.


## Conclusion
Every business, no matter how big or little, needs a digital platform as we become more dependent on technology. And one of the most advantageous platforms at the moment is a web application.
By hiring the Communere Ltd team, you can ensure your web app development project will succeed. Communere developers with state-of-the-art technical proficiency deliver the best result at an affordable price.

This highly experienced and dedicated team can develop a business-specific application for you or your organisation. For over a decade, they have developed high-performance, feature-rich mobile and web applications that can be operated on all major platforms, including Android and iOS.
The Communere developers use the most advanced management methods to ensure that your order will be delivered on time.

In addition to developing mobile applications, Communere Ltd. has extensive experience in implementing business-oriented software.
By utilising a variety of technologies, they develop high-performance, feature-rich web applications as well as mobile applications for all major platforms.
Moreover, the company offers simple technical solutions to business challenges through its own front-end and back-end development teams, all of whom have extensive experience in online technology.

What Is A Web App and Why Your Business Needs One?

<figure>
<img src="https://content.communere.com/assets/0115cb36-0bcf-4de3-955c-6a51b8ffa8d6" />
 <figcaption style="text-align: center;">
</figcaption>
</figure>
# Introduction

The task of finding the shortest path between two nodes in a graph is well-studied in computer science literature. The well-known Dijkstra algorithm finds the shortest path from a source to all other destinations. In this post, we explain how did we handle routing in the public transportation network for one of our clients at Communere.

## Public transit network

Public transportation differs from other networks. The main reason is that the edges are available only at certain times. For example, consider a bus that leaves stop X at 08:00. The edge between stop X and the next stop is available only at 08:00. This proves to make routing much more difficult. The following questions must be considered in this scenario.

- How long should we wait at a stop?
- How to handle the dynamic nature of the network?
- Will there be any path from my next stop?

These problems make traditional routing algorithms inefficient in this case.

## RAPTOR algorithm

Round-Based Public Transit Routing (RAPTOR) is a temporal pathfinding algorithm developed by Microsoft. The original publication of the algorithm by Delling, et al. is available [here](https://www.microsoft.com/en-us/research/wp-content/uploads/2012/01/raptor_alenex.pdf). This algorithm was developed to address the mentioned problems. It is widely different from Dijkstra. It doesn't even need a Min-Heap to prioritize the shortest paths!

## Algorithm Intuition

In public transit network, each means of transportation moves along its predefined route. Unlike general-purpose routing algorithms like Dijkstra, RAPTOR utilizes this property to find the optimal path. To illustrate the difference, consider the following scenario.

The graph below depicts a public transit network and its availability times. Each colored line represents a path that a bus takes. The red line starts from stop 1 at 08:00 and arrives at stop 7 at 10:30. Stops 2, 3, 4, and 5 are the intermediate stops in this route. Each intermediate stop is 30 minutes away from its neighboring stops in the route.

Assume that we want to find the earliest arrival time for stop 7 given that we are at stop 1 at 7:45.
![public transportation network](https://www.dropbox.com/s/86hivofa9aq0zi3/trips.png?dl=1)
The first thing would be to find the earliest leaving bus (or other means of transportation). In this case, there is only one bus leaving at 08:00 so we have a total waiting time of 15 minutes at stop 1.

Once we catch a bus, the algorithm will traverse that route to the end. Put simply, once you hop on the bus at stop 1, the algorithm will sequentially traverse all the stops that are on that route. And, it will record the arrival time for each stop. So, after traversing the first route, the algorithm will have the following records.

- Best arrival time for stop 1: 08:00
- Best arrival time for stop 2: 08:30
- Best arrival time for stop 3: 09:00
- Best arrival time for stop 4: 09:30
- Best arrival time for stop 5: 10:00
- Best arrival time for stop 7: 10:30

With this information, the algorithm can stop and report the earliest arrival time for the desired location is 10:30. This is a valid answer if the desired number of interchanges is 0. However, there may be a better path by changing lines.

The algorithm works iteratively. Routes with 0 interchange are considered in the first iteration. The second iteration builds upon the knowledge obtained from the first iteration and considers routes with 1 interchange. More specifically, from the first iteration, the algorithm knows that the earliest arrival time of stop 2 is 08:30. Therefore, it starts the pathfinding routing the same way as before. The only exception is that it starts from stop 2 at time 08:30.

To clarify things a little more let's trace the second iteration of the algorithm. In the first iteration, the algorithm marks any stop that has been improved in terms of arrival time. Since each stop's arrival time is initially set to infinity, all stops in the first iteration are improved. In the second iteration, the algorithm considers the available trips from stops 2, 3, 4, 5, and 7. Let's trace the blue trip from stop 2. Given that we arrive at stop 2 at 08:30, we can board the blue route and reach stops 6 and 5. The respective time for arrival is 08:40 and 09:00. As you can see, the arrival time for stop 5 improves. After the second iteration, the best arrival time for each stop is as follows.

- Best arrival time for stop 1: 08:00
- Best arrival time for stop 2: 08:30
- Best arrival time for stop 3: 09:00
- Best arrival time for stop 4: 09:30
- Best arrival time for stop 5: 09:00 (follow the red line to stop 2, then switch to the blue line)
- Best arrival time for stop 6: 08:40 (follow the red line to stop 2, then switch to the blue line)
- Best arrival time for stop 7: 09:30 (follow the red line to stop 3, then switch to the green line)

The iteration continues until no improvement can be made or the algorithm hits the maximum number of iterations.

Another important aspect is the ability to walk between stops. RAPTOR does not compute the walking time between stops. This is a brilliant abstraction as any routing algorithm can be used to find the walking times.

### Performance

RAPTOR has become the de-facto algorithm for public transit routing. Most of the other algorithms need preprocessing stage to have an acceptable performance; this is not the case for RAPTOR. Most of the queries are almost instant in small networks. This algorithm proved to be more than three times faster than the modified versions of Dijkstra.

### Footnotes

- The original paper is vague in some aspects. And, it also has minor bugs in its pseudo code. If you are stuck, [this link](https://ljn.io/posts/raptor-journey-planning-algorithm) provides a better explanation.
- There is a python implementation available [here](https://github.com/lmeulen/pyraptor).

Routing in the public transit network

<figure>
<img src="https://content.communere.com/assets/9f8b443b-166e-4f62-bf31-b12e7d0a30d5" />
 <figcaption style="text-align: center;">
</figcaption>
</figure>

A gentle introduction to operation research

## Motivation

The demand for data analysts has exploded in recent years. Most companies are integrating data into their decision making process. However, in many practical applications, the company may not have data to answer the questions. An example of this scenario is when the managers want to start building a new product. In this case, many questions must be answered before the production. Some of these include

- How many units should we produce?
- Is this product profitable?
- How much budget should be allocated to marketing?

As you can see, these questions are directed toward areas where previous data may not help. So, should we just trust the intuition of the managers or is there a more systematic way to explore the answers? In this post, we explain the latter.

## Mathematical Programming

Most of the decisions in a firm are either trying to maximize profit or minimize cost. In a way, the decisions are trying to achieve one of these extremes. You may remember this graph from high school; and, probably the “pointless” differentials.

<img src="https://content.communere.com/assets/b5a2fee2-9266-4c2c-aad1-a34b673b037f.jpeg" />

Even though Y = X2 does not correspond to any particular thing, it can represent a whole range of possibilities. For example, if we substitute “Y” with “Production Cost”, things start to make sense. Now, trying to find the minimum value of “Y” is not a math exercise anymore; it is the process of minimizing the “Production Cost” given the parameter “X”. Parameters “X” and “Y” can represent anything. This is the simplest case of decision making where we control parameter “X” and the value of “Y” is derived from our decision. In this example, the relation between these parameters is quadratic. This is the essence of operations research and mathematical optimization.

### An optimization task is defined by

- #### **Decision variables:** 
The parameters that you can control.
- #### **Objective function:** 
The thing we are trying to optimize (either by finding the minimum or the maximum) expressed in terms of decision variables.
- #### **Constraints:** 
A set of real world constraints. For example, the maximum budget.
And, that’s it! Just by defining your problem and constraints in mathematical terms, you can have the solution using Microsoft Excel!

### Real World Example

Suppose you want to start a business. You have 10000$ and want to build your awesome idea. You can buy 1 unit of material that is needed to build your idea for 2$. And, buying 100 units at once has a 10% discount. However, you can only work 4 hours a day because you have a full time job. But, you can always pay others to work for you. In this case you can pay someone 50$ per hour to work for you. Each hour of work produces 3 units of your awesome idea which you can sell each for 12$. And, each unit requires 2 materials.

Given these terms and a deadline of 30 days, how would you allocate your budget to maximize the profit?

### Let’s translate this description into our framework.

- #### **Decision Variables:** 
There are several parameters that we control.
 - Number of raw materials to buy?
 - How many hours should we recruit someone to work for us?
 - How many hours per day should we work?
 - How many units to produce?
- #### **Objective function:** 
Maximizing profit in terms of the costs associated with each decision variable.
- #### **Constraints:** 
There are several constraints.
 - Our budget.
 - Each unit needs 2 materials to be produced.
 - You can only work 4 hours per day.
 - A 30 days deadline.

As you can see, there isn’t any formula or equation in the new form. But, the beauty of this framework is that you have to think about the situation explicitly. Every cost and its associated profit is now explicit. Even if you don’t want to enter the realm of optimization, you have a clear idea of what you can expect. The objective function tells us that we do not control the profit directly. It is our decisions in resource allocation that determine it. Furthermore, we can do some “What-If” analysis. For example, how much money can we make if we don't outsource the production?

## Applications

As a company grows, the constraints for its decisions grow too. For example in the case of public transportation planning, hundreds of constraints must be taken into account.

As mentioned previously, even if you do not intend to use the mathematical framework for optimization, having the problem in a concise and explicit way is an invaluable tool. However, finding the optimal solution using computer softwares like Excel or IBM CPLEX opens the door for in depth analysis and can influence the decisions.

Mathematical optimization for decision making

<figure>
<img src="https://content.communere.com/assets/c6d40ef8-791e-4ce3-8a24-2c638b159c56" />
 <figcaption style="text-align: center;">
</figcaption>

If you're a business owner, you know the importance of investing in technology. Nowadays, most companies are going digital and moving their operations online. In this article, we will cover everything that you need to know about **web applications** and why your business should invest in them!

## What is a web app?

Web apps are a type of web application that runs on your computer or mobile device. They're
hosted on a remote server, so you don't need to install anything on your local machine.

This means that when you open the app in your browser (for example, Google Chrome), you will
have access to all of its features without having to download any software.

### How to Create a Web App
 
Web apps can be written using any programming language, though some languages work better than others depending on what kind of functionality they offer and how much control over performance they give developers (for example Java versus Python).


### What makes web apps unique
 
Web apps are unique in that they are hosted on a server and accessed through a URL (e.g., [website]/app). This means that they can offer much greater security than desktop apps because they run through a browser and have access to fewer resources than traditional desktop apps. These apps can be accessed from any device with an internet connection!


### where do web apps run
Web applications run through servers hosted either by yourself or a third-party vendor such as Amazon Web Services (AWS) rather than directly installed onto individual devices like desktop apps; however, they still require some sort of internet connection due to their reliance upon cloud-based resources. As the data is stored remotely via backend databases (rather than on end users' devices), they tend to load much faster and offer greater security.

## Web app uses
You can use web applications for a wide variety of uses: from a giant corporation to an individual, for several purposes. Most used web apps are webmails, online image editing, E-commerce shops, etc. you can access web apps by any browser, although specific browsers can access some web apps.

## How web apps work
 
For using web applications, you don't need to download any apps since they are accessed through a network. Users can access a web app via a browser like Google Chrome, Mozilla Firefox or Safari.
You may need a Web server, application server, and database to operate a web application. Web servers take care of the requests that come from a client while the application server completes the requested task. The database can be used to store any required piece of information. 
Fortunately, web applications have short development cycles and can be completed through a small development team. Web developers mostly use JavaScript, HTML5, or CSS to build a web app.

## Benefits of a custom web app
Web applications are the easiest to update and maintain, which leads to a faster product launch. We can name so many advantages of the web app, but the main ones are:
- you don't need to install web apps.
- they can easily be accessed through multiple browsers.
- web apps can be accessed on different devices, including desktops, laptops, or mobile.
- Easy maintenance: they have a common codebase regardless of the mobile platform.
- you have no update hassle.
A well-designed and built web application can deliver a huge amount of value—and it doesn't have to cost you an arm and a leg either!

## Desktop apps - what you need to know

Desktop apps are here to stay. A web app is in no way a replacement for a desktop app, just as a desktop app is not a replacement for a web app. Desktop, web and mobile apps all form a delicate ecosystem of applications that each have their own place within the world of software development. Each type of application serves a different purpose, however. No one solution will work for every business case or industry.

Desktop applications are designed to be used on personal computers that use Microsoft Windows or Apple Mac OS X operating systems (Windows PCs and Macs). They run locally on your device, usually without needing any type of connection to the internet. Desktop applications can do everything from handling email accounts to building complex databases with advanced search functions—they're built from scratch based on how you want them to function within your business workflow and what tools you need at your fingertips when accessing them regularly.

## The rise of web apps

Web apps have been steadily rising in popularity over the past decade and for good reason. Web apps are easier to use, more accessible, and offer a better user experience than traditional software. They're also more secure, and cheaper to build and maintain, deploy, update and integrate into your business system.

## Conclusion

In short, web apps are an increasingly popular option for businesses that want to build a customised application. And the good news is that they're also great value for money. With just a little bit of planning and preparation, you can create something truly amazing - which will pay dividends in the long run! 

Wherever you are in your journey towards launching your web app, Communere Ltd. hope this article has given you some useful insight into what it can do for your business. Regardless of where you are in your journey, Communere's experienced team can help you!
[Get in touch with us here.](https://communere.com/contact)

Why should your business invest in web apps?

<figure>
<img src="https://content.communere.com/assets/b25b9415-406e-4c5e-8acd-1033a2804fe3.jpg" />
 <figcaption style="text-align: center;">Photo by Luke Peters on Unsplash
</figcaption>
</figure>

It is a year since the UK went into lockdown, and businesses are continuing to flourish on a remote basis.

IT leaders quickly adopted a work from home strategy that enabled them to not only support the BAU operations of an IT function, but also to carry on with huge development projects.

From the conversations we are having with senior IT decision-makers, most are claiming the shift to home working has been successful. However, they are quick to ask us for our **top tips for running a remote IT function.**

So that got us thinking, how do we run an efficient remote team at Communere, and how do some of our clients do it so productively?

- We run what we call a ‘virtual office‘. Utilising all Slack features, our developers can communicate and chat spontaneously as though they were in the office. We actively encourage this **constant conversation**, and we feel it goes along well in cementing our company culture.
- We regularly check on our **IT teams home setups** and provide facilities for them to upgrade their laptops, monitors, keyboards, mouses, chairs and desks to the most suitable and ergonomic ones.
- Many clients also tell us how they have mapped out a **training budget** to ensure their staff continue to progress, even if on a remote basis. We love this at Communere. We feel this is an absolute must to ensure you retain top talent. Our **weekly educational/technology** meetings help our team stay tuned to the emerging technologies.
- To keep the fun part of the physical office alive, we introduced our **weekly remote tea party**. Our team gather remotely and enjoy talking about their hobbies, such as music, football, films.
- **Team yoga**, get the whole IT team from 1st line to the Dev team joining in. It would make for some fabulous team photos, if nothing else.
- **Virtual coffees**, we have always been remote and will, of course, continue to remain remote. We have regular check-ups with our developers to see how things are progressing and if they need any additional support to hit our clients’ deadlines.
- We are not for micro-management, but we do feel as though **goals need to be clearly outlined** when teams are working remotely. A team goal will mean everyone will have something to aim toward; this should encourage teamwork.
- Breaks are important! **Encourage** the team to take short and **regular breaks.**

If you would like a discussion about how a remote development team could help your business, feel free to let us know. We would be happy to explain the benefits of using us via **virtual coffee.**

A Successful Remote IT Team - Communere's Approach

<figure>
<img src="https://content.communere.com/assets/74b8ae31-bb10-4538-845c-b1cf1ad3b485.png" />
 <figcaption style="text-align: center;">source: unsplash</figcaption>
</figure>

Starting a business? Communere’s Step-by-Step Guide to Your Starting Your Company

You are walking to work and suddenly, a genius business idea pops into your mind. Or you have always dreamt of starting a business, although you have not had the time and resources, but now is different. Are you looking to turn the idea into a reality; here is a guide to those first steps you should take.

It can feel like a daunting prospect setting up a business, it can be difficult to know where to start and there are lots of questions i.e. do you already have a network in place, who are competition, where will you receive funding etc. Taking small steps will help with the bigger picture questions.

## An easy-to-follow guide has been put together by Communere:

1. ### **Know the market:**
 Ensuring that there is demand for your idea or product will help to ensure that you are looking in the right place to starting a business. Have you spoken with your prospective customers; do they feel your idea could benefit them in some way? Having a plan before you start will help to ensure there is a gap in the market (or a space for a new gap to be made) will help to alleviate any future. Here is a [great article](https://blog.nationwide.com/market-research-business-ideas/) to show the types of market research that you will need to do before you launch into things.

2. ### **Figure out how to structure your business:**
 Making sure that you are set up as the correct business type is important and there are pros and cons to each i.e. becoming a sole trader, limited company or LLP. Setting up a Limited Company is a big commitment, you need to be fully aware of the annual tax and filings responsibilities that you will have to adhere to, the responsibilities have been clearly laid out here. There are some very useful guides to which type of company you should set up depending on your size and plans for growth. [Here](https://www.gov.uk/running-a-limited-company) is a great article to help you establish which type of business you should consider opening.

3. ### **Making your business plan:**
 Building a business is similar to building a house, you need to plan as if you are drawing a blueprint for a property and start from there. A clear, concise business plan will help by enabling you to envisage where the business will be in 5 years, makes a great reference point for growth to look at in the future, and it also helps investors to take your business seriously if investment is a key stage of your business plan.

 If you need some advice on how to draw up a business plan, bgateway have a [free template](https://www.bgateway.com/resources/business-plan-template) available for people to use which helps in those early stages.

4. ### **The name; choosing a name guide:**
 Choosing a good name for a business can be the difference between “make or break”, for the company. The name needs to be professional, memorable and simple. [Forbes](https://www.forbes.com/sites/allbusiness/2016/10/23/12-tips-for-naming-your-startup-business/#72fa6a27904e) have put together a brilliantly insightful list about how to name a business from ensuring that you are not limiting yourself as a business when you grow, to making sure that the name conveys meaning.

5. ### **Logo designing:**
 The logo can be a big window into the quality of a business from a consumer’s perspective, it is the window to business soul as this is the first thing they will see along with the name. Today, there are so many businesses that are equipped to help you with the logo design and ensuring that they are original. [Tailor brands](https://www.tailorbrands.com/logo-maker) offer several different packages to help with logo design and they also offer website building services too.

6. **Website design:**
 Once you are happy with the name, logo and want to proceed with the business, now is the perfect time to build your website and start attracting customers to your amazing new business venture.

Most importantly when building a business, relax and enjoy the process. There is so much to be gained from owning your own company and the rewards far outweigh the challenges. We hope you enjoyed Communere’s first building a business series, we will be back soon with more business building advice.

Step-by-Step Guide to Start Your Business

Despite it being early in the year, we are already set for a big change over the next year to come. People working remotely is going to be on the rise and organisations are fearful of the challenges ahead for their employees. Some of the biggest fears from people at this moment in time is how to keep employees engaged and keeping them learning while working from their home office.

Understandably, with social distancing being enforced at this stage in time, most businesses must adapt to remote working in order to succeed in this tricky climate. Businesses need remote employees now in order to be able to succeed, although that doesn’t guarantee success.

## Keeping People Connected

Going from working in a busy and buzzing workspace to working in an office space alone can either be the most exciting challenge or something people can dread. Loneliness can be a serious challenge for transition for people and empathy from employers is vital here. 

Positivity, trust and keeping motivation high is an employer’s responsibility and is important while keeping remote employees happy.

### How Does Working Remotely Affect People
Remote working can cause anxiety or stress for some people, which can be a huge hindrance when it comes to productivity.

### How To Keep a Remote Team Engaged
Choosing a business tool that keeps employees engaged with each and encourages team building is important right now. Games can help with engagement and this can be a nice way to keep things light at times, the best tools that we have seen are:

- Team Building
- Virtual Meetings with games such as bingo
- Team Building with Donut on Slack
- Coffee Meeting

## Keep People Learning

Encouraging learning among employees who are working remotely will provide them with a goal to achieve. This will also help them to grow as an individual, becoming both a better employee and a more fulfilled individual. If someone is able to work an hour before work or at the end of the day (when their usual commute could be), yet at the same time learn a new skill such as something technical, marketing related, HR, language, or sales related. This is beneficial for both them and the organisation, all employers should encourage learning at all, although especially now when morale is low.

### How to Encourage Your Remote Team For Learning
Here is a list of companies that help to encourage learning among your remote workforce, they offer affordable, short and long courses:

- #### Udemy 
  choose from thousands of short courses to suit your employee’s skillset/ development area
- #### Coursera
  courses, certificates and degrees, choose from lots of different courses and professional certificates
- #### EdX 
  Free online courses offered from Harvard, MIT and more of the worlds leading organisations

## Preparing for the future

Finally, with recent events in mind, the world of work will restore to some level of normality as it has done previously. Nobody can predict how long these changes will last, although with a motivated and educated workforce you can guarantee they’ll be happy. This can be achieved with continuous team building and games. And when things do finally go back to normal, you will have a newly educated team to go back to that loves your organisation.


Keep Your Workforce Engaged and Educated Who Working Remotely

<figure>
<img src="https://content.communere.com/assets/d8355a6a-9866-41e5-8e4b-f526a715aa5d.jpg" />
 <figcaption style="text-align: center;">source: unsplash</figcaption>
</figure>

Achieving peak performance during quarantine at work is going to be a challenge for most people, we are used to clear targets and busy work environments.

We have put together a short guide to working effectively, while looking after your mental health during social distancing or self-isolation during the next weeks or months.

## 7am – 8am Yoga or Meditation

Ensure that you are starting the day in a peaceful and calm way. Set intentions for the day as to what you wish to achieve and try your best to stick with them!

## 8am - Breakfast

Stick to something healthy; with food shortages now, it can be difficult to get hold of fresh fruits, a great way to get in the vitamins and minerals needed is to buy frozen fruits. Frozen fruit in porridge or granola and heated fruits are a perfect way to feel fresh and ready after your yoga session.

## 9am – Start Work

Ensuring you have a workspace that is not connected to your “normal” day to day household activities is important in order to be able to disconnect from work in the evenings. Ideally this would be an office with its own desk.

A dining table does also work, although make sure to clear your laptop away in the evening so that home life remains just that, home life.

## 11am to 11.15am – Go for a Walk

Get outside! With Spring on the way, there will be hopefully plenty of sun to brighten our days over the coming weeks/ months.
The most important thing to remember is to get some Vit D. For most people, commuting provides a small amount of their daily walking/ exercise.

This is a great time to walk to the end of the road and back again to boost productivity and making sure you still get steps in. Of course, while self-isolating/ social distancing, remember to stay 2 metres away from everyone else.

## 1pm – Run

A quick run to boost your immunity and keep your focus throughout the rest of the afternoon is an ideal way to spend half of your lunch break.

## 1.30pm - Lunch

Keeping with the theme of a healthy lifestyle; lunch should be a mix of at least 3 different types of vegetables and wholegrain foods such as brown rice, wholegrain pasta or chickpea pasta. Our immune systems need as much goodness as possible in their diet currently. Some other lunchtime alternatives:

- Vegetable Soups & wholegrain bread. Make a batch of a vegetable soup on Sunday evening and make enough to last the week.
- Rice noodles and vegetables with meat or tofu of your choice. Keep it colourful and add lots of fresh vegetables
- Vegetables & couscous with sauce. Again, a super cheap and healthy way to boost your immunity during these tricky times.

## 2pm-5pm Work Tasks

Meetings galore and keeping in touch with management, ensuring all things are tied up and finished off so that you can enjoy your evening.

## 6pm Dinner & Evening Time

Family dinner time to rest and unwind, spending time together as a family is vital right now. Some great easy dinner ideas can be found here at Jamie’s 15 Minute Meals. These are healthy and will be great to get kids involved in making too!

This healthy guide will enable you to save money every day, boost your immunity and keep you from feeling like you’re working all day everyday from home. Hope this helps!

How to Work Efficiently During Quarantine

<figure>
<img src="https://content.communere.com/assets/edfcee36-259a-42ef-8280-296c33b3a316.png" />
 <figcaption style="text-align: center;">source: unsplash</figcaption>
</figure>

In the wake of a coronavirus outbreak, global warming and the numerous benefits to having a remote workforce. Businesses are having to adapt their working policies and encourage employees to work from the home office far quicker than previously thought.

The rise in remote working will be increasing significantly over the next few months, with the outbreak of coronavirus increasing, companies are needing to adapt their workforce to work from home, for some organisation they’ve had time to adapt to these changes. Although for most companies, this is a completely new era for their organisation which could bring many benefits and challenges. Similarly, for employees, working remotely could be everything they have hoped for, although it could also be a transition period that bring feelings of isolation and irritability.

## Create Daily Routines

The idea of working remotely can be a dream come true; not having to face the stressful commute, reducing monthly outgoings on travel and avoiding people with the stresses of coronavirus being prominent now. However, some people can struggle with the loneliness of working from home if they enjoy the vibrant office atmosphere and chit chat that many people enjoy. There could be a lack of motivation if meetings are a big part of someone’s role also, ensuring all these weekly meetings go ahead is vital to ensure employee engagement and maintaining a motivated workforce.

It is important for companies to implement daily routines, ensuring the Monday morning meeting goes ahead at the same time every week, or the daily Friday catch up stays the same. A daily report can be written, or a project management tool can help to ensure that businesses stay in the loop, while maintaining that trust of their valuable employees. Ensuring that tasks are completed can be the new shift for organisations as remote employees may prefer a different time structure; they could choose to work earlier and finish earlier or start later and finish later, allowing employees to be target approached will ensure that they feel supported and trusted in this new shift.

## Invest in Good Systems

One of the major concerns for organisations is ensuring that employees continue to work effectively while working from home. Micromanagement plays no part in an employer need to view results from their remote workforce; ideally it would work in the same way as in an office with a weekly catch up and targets set for the week.

At Communere, as a fully remote workforce, we are still learning the different tools that work for the different departments, being an international team, we need to feel like a “normal” office while retaining our remote status. We use a number of effective systems to help keep everything ticking over; Jira for daily updates, Confluence for weekly meetings and targets, as well as Zoho for our CRM system to ensure sales targets are being reached, and for most communication online and for calls we use Slack and Skype.

Technology and tools play a huge role in allowing for employee targets to be reached and ensuring they remain motivated and comfortable with what they need to achieve. This in turn leads to happy employees, which is vital for all businesses for productivity and retention of staff.

## Offer Training & Development

One mistake organisations can make while working with a remote workforce is that they hire for the most skilled in the market to ensure that there is limited training time, although once people are in the organisation, there can be a lack of support and development. One of the biggest factors of ensuring employee retention is to provide opportunities such as online courses, live webinars and conferences for them to attend.

The cost of these courses, conferences etc is an investment for your organisation as it ensures loyalty, networking opportunities and time for team bonding; all vital things for employees in both the office and remote working space.

## Trust, Trust & Trust Again

Of course, organisations may fear the worst with a remote employee; sat at the pub enjoying the sunshine or taking a week of holiday while occasionally checking in. This could happen with office-based employees and is not exclusive to home working employees. The targets of deliverables set by you, ensures that targets are met, and employees feel trusted. If employees feel trusted, they thrive.

With these tips in mind, we hope you can thrive as an organisation, or as a remote employee both now during trickier times, and in the future if it is something that you enjoy. At Communere we thrive as a remote team and feel free to ask for any tips. We are happy to help.

Guide to Remote Working

<figure>
<img src="https://content.communere.com/assets/c00d3bfc-d308-45df-a72f-3a0a5d01dc17.png" />
 <figcaption style="text-align: center;">source: unsplash</figcaption>
</figure>

Communere are a truly remote workforce, we have daily meetings and catch ups throughout the day and week, although the remote nature of our business allows us to be truly global. It is estimated that by 2029, 38% of the workforce around the world will be remote working. We will explore some of the benefits to remote working and the benefits to Communere.

## Promotes a positive work-life balance

A work-life balance is a benefit that a lot of employees value as a top priority when choosing the best place to work. The ability to be able to take and pick up children from school, going to a fitness class early or at lunch time, going to the doctors or making it to the supermarket before the mad evening rush, or just the ability to be home when for an important delivery. These are all things that make life a lot easier, and are all possible when working remotely. These may sound like small things, although it makes a huge amount of difference in the long run.

## Increased Productivity

This can be one of the biggest fears for an organisation while transitioning to a remote workforce; how can we tell that our employees are working hard if we can’t see them? According to a study conducted by Stanford University, most organisations that have adopted remote working have found that there has been an increase in productivity. Without the many distractions, remote workers can focus entirely on their daily tasks, allowing for quiet time, people can focus entirely on their projects. Thus, increasing productivity in the long run.

## Money Saving (for the organisation)

Many remote workers use their own equipment, this is Wi-Fi, heating, among others which have a big impact in the long run for saving money as an organisation. The company does not need to rent large, expensive offices; this is one of the biggest savings an organisation can make over a long period of time.

Finding the best talent globally with ample experience is a much easier task when you are looking for remote employees. Most remote employees have years of experience in an office. Therefore, the upskilling time is reduced which also saves companies a large amount of money in the long run. The Global Workforce has a table of savings for an organisation per year for a remote emplyee:

- Circa £1,857 is saved per remote employee in reduced office sizes and utility bills
- £3,560 is saved per remote employee by increasing productivity
- £464 per remote employee by decreasing turnover through increased flexibility and happiness
- £681 per remote employee due to the fact that remote workers are less likely to take sick days
 Total savings: £6562.00 per year per employee

## Money Saving (for employees)

The money saving doesn’t stop with the employers, working remotely can save individuals thousands each year. According to Global Workplace Analytics, people who work from home half time can save between £1,500 to £5,000 a year, that is double for full time remote employees. The savings for employees can be found on petrol, the running of their cars on a daily basis, rail or bus travel, parking. There is also a reduction in daily coffee spending i.e. the daily Starbucks on the way to work, as well as lunches bought in supermarkets etc. Big savings can be found in child care if people are able to pick up their children this can reduce childcare costs, as well as the savings on office attire. Over time, these savings can be significant and allow for a healthier, happier work life balance.

## Good for the Environment

It is vital for organisations to reduce their carbon footprint. According to the US environmental protection agency a “a typical passenger vehicle emits about 4.6 metric tons of carbon dioxide per year. This assumes the average gasoline vehicle on the road today has a fuel economy of about 22.0 miles per gallon and drives around 11,500 miles per year. Every gallon of gasoline burned creates about 8,887 grams of CO2”.Remote workers on the other hand are not responsible for adding to the daily carbon emissions by car commuting, which could be up to 30 miles per day on average, saving up to 7,830 miles on a cars emissions for an average working year of 261 days (depending on how much they use their car of course). You can track your daily emissions with Map my Emissions. Working remotely can help:

- Less traffic pollution caused by fewer employees driving to work.
- People being less inclined to purchasing throwaway cups or containers on breaks.
- Conserving office space.

Being in control of factors that can damage the environment (such as office supplies and the use of lights)

Remote employees are less inclined to use one use plastics and paper cups for their morning coffees, as well as consume less food that comes in plastic containers, with plastic bags among others.

## How Communere’s Remote Team Can Help Your Business

Our team are based in different locations around the world and they are all working on different projects, although we all have one thing in common; our love for technology, remote working (and Communere of course!). The benefits for us an organisation are endless:

- Employee retention: we have a loyal workforce that are given a lot of autonomy, we have our daily catch up calls, although they are trusted to work on their projects and most importantly, they always deliver.
- Agility: the remote nature of our workforce allows for time to be spent on projects, as there is no daily commute, we are rested and ready to work on your projects
- Low Carbon footprint: we care about the environment and working remotely contributes positively

The benefits of working remotely are endless, and with the stark increases in productivity, reduction in carbon footprint and endless benefits, it is no wonder companies and individuals are opting for the remote working or part remote working lifestyle.

Remote Working Benefits – Why Communere is Ahead of the Curve, And Why This Will Benefit Your Organisation.

A legacy system is a computing software / hardware system that a company uses, despite it being outdated. The system is still being used by a company, although it still meets the needs of what the company designed it for, it is not a scalable system that enables growth both with, and for the organisation. The system will continue to do what it has always done, although it will never do any more than it was originally tasked to do. The downfall of a legacy system is that it cannot interact efficiently with new systems such as the cloud.

Technological advancements over recent years have led the majority of companies to have major issues with existing legacy systems. In order to scale and reduce costs, companies need to adapt new technology; although this comes at a price. For companies to keep aligned with the latest technical capabilities, companies need to move away from keeping their legacy system and move to cloud and computing and software migration.

Moving a business through a software migration can be a time consuming and costly process, here we will look at why moving from a legacy system to a software migration is vital for a companies success in 2020.

## Why a Software Migration Strategy is Important

There are multiple reasons a business may be looking to having a software migration; this could be to outperform the competition, or a more agile and efficient internal system. Whatever the reason a company is adopting a new system; the process needs to work.

When looking at an unsuccessful migration, the reasons for this can be from human error; where the inputted data or code is not accurate which can lead to major problems for an organisation. Although this can also when everything has gone according to plan during a software migration also. Most worryingly for a company is when there were no errors within the original software code, or data, and errors present themselves when the code has been inputted into a new, more efficient system.

For example, a system that has been developed with the previous version of .NET or ASP classic, Now we can help to migrate the software and build it from scratch with the latest and cutting-edge technologies e.g. ASP.NET Core which can be run on all OS(s) and React.js systems.

The best way to avoid any major issues within a software migration is to create a software migration strategy for the predicted duration of the project on a week by week basis. Working in agile is of course the most efficient process that a company can use for this process to be a success. The biggest challenges for a software migration strategy can be budgetary, a lack of clear strategy and working in an outdated methodology. The best way to ensure that a strategy is executed in the best way possible is to ensure that there is a dedicated team working full time on the projects until completion. At Communere, we have a full dedicated team of remote employees to help an organisation with their software migration strategy.

Communere work to a tailored strategy for organisation, although it looks something like this:

- Understanding the software Before an effective migration, a complete analysis of the legacy software that has been used, and the software platform that you will be moving to is completed to ensure you're receiving a tailored software plan.
- Cleanup The source software is checked for anomalies, cleaned up in order to be able to effectively start the software migration plan.

- Continuous Maintenance Software regularly has new bugs (even with the best code in the world!) which needs regular maintenance to ensure your product works efficiently and most importantly, that customers data is safe

- Best Practice & Governance We follow the latest international standards for best practice and governance to ensure a well built and efficient product that adheres to high quality standards.

## Best Practices for Software Migration

- Back up the software before executing. The most important part of the process. Ensuring the software has been backed up before starting.

- Keep strategy. Ensuring that the strategy is upheld throughout the process to ensure the project runs smoothly will help both the cost and time of the migration

- Test the Software. Throughout the process of the software migration, ensure that the code is being tested as you go.

## Migrating Software to the Cloud

Ensuring that your dedicated developer is skilled in uploading the code to the cloud will protect your companies code and data in the long run. The cloud has many major benefits for an organisation allowing scalability and growth. Communere is equipped to enable your software migration process.


Communere Steps Guide Software Migration for Your Organisation

<figure>
<img src="https://content.communere.com/assets/4c19084d-7149-4dd4-84c7-c919163ef55f.jpg" />
 <figcaption style="text-align: center;">source: unsplash</figcaption>
</figure>

Encoding Categorical features is one of the fundamental feature engineerings processes when using machine learning techniques. There are different techniques and packages to perform Encoding, but it can be not very clear which one to use or what exactly their differences are.
Here, I am going to give a quick review of most common encoding techniques and how to apply them on the data.
The dataset, I am using to demonstrate, is Titanic dataset available on Kaggle.

## Outline

- OneHot Encoding
- Lable Encodeing
- Get_dummies
- DictVectorizer
- OrdinalEncoder
- FeatureHashing

### OneHot Encoding

Provided by **Scikit-learn** library, it utilises the one-hot encoding (dummy) scheme. This transformer needs array-like of integers or strings as an input. The output will be a sparse matrix or an array with binary columns for each value of the categorical features.

We need to extract the categorical features and data from the dataset. Then, instantiate OneHotEncoder and fit and transform the data into the OneHotEncoder to transform the categorical data into 0/1 code.

In this example `df_categorical_data` contains 2 columns namely `'Sex'` and `'Deck'` where `'Sex'` can have female or male and Deck can have A, B, C, D, E, F, HA value respectively.

```
# import OneHotEncoder
from sklearn.preprocessing import OneHotEncoder
# instantiate OneHotEncoder
ohe = OneHotEncoder(categories='auto', sparse=False )

# apply OneHotEncoder on categorical feature columns
Categorical_ohe = ohe.fit_transform(df_categorical_data) # It returns an numpy array
print(Categorical_ohe.shape)
Categorical_ohe
```

<figure>
<img src="https://content.communere.com/assets/202a905e-2ac4-4d9c-985c-e994099f7ddc.JPG" />

 <figcaption style="text-align: center;"></figcaption>
</figure>

The output is an **array** with 9 columns covering all possible values that `'Sex'` and `'Deck'` can get.
We can retrive the name of the columns and convert the array to a dataframe if necesserly by utilising `get_feature_names()` attribute of the Onehot Encoder and DataFrame method of Panda API as follows:

```
cols = ohe.get_feature_names() #get the column names using .get_feature_names() attribute.
df_Categorical_ohe = pd.DataFrame(Categorical_ohe, columns=cols) #convert the result matrix to datafarem (if necessary)
```

<figure>
<img src="https://content.communere.com/assets/8574d067-45b9-4c83-abfe-aa5f2d3f0e38.JPG" />

 <figcaption style="text-align: center;"> </figcaption>
</figure>

Note that when we use One-hot Encoding on a single column DataFrame, we need to let Numpy reshapes the DataFrame to make it compatible with the input format.

```
# apply OneHotEncoder on one categorical feature
Categorical_ohe1 = ohe.fit_transform(df['Embarked'].values.reshape(-1, 1)) # It returns an numpy array
# by using reshape(-1, 1), we indicate that we have 1 column but rows are unknown and we want that numpy figures the rows out such that it is compatible with the original shape.
Categorical_ohe1
```

<figure>
<img src="https://content.communere.com/assets/3157671c-e7f7-4be6-8bae-63f9b3b8a54f.JPG" />

 <figcaption style="text-align: center;"> </figcaption>
</figure>

### Lable Encodeing

Provided by **Scikit-learn** library, this class of transformer can be used to transform numerical and non-numerical labels into numerical labels. In contrast with one-hot Encoder, which provides a one-hot coding scheme (‘one-of-K’), LableEncoder provides labels with values between 0 and n_classes-1. It is advised to use this Encoder for target values rather than input values.
The output type of lableEncoder is similar to its input, but you are restricted to the [array-like](https://scikit-learn.org/stable/glossary.html) of shape(n_sample,). So, in case, you want to encode a DataFrame with 2 columns you should first convert it to an array-like of shape(n_sample,), perhaps using `raval`. I provide a couple of different scenario in the followings:

```
from sklearn.preprocessing import LabelEncoder

# instantiate a LabelEncoder object
le = LabelEncoder()

# feed a coulmn of the datafarme which has a binary class feature i.e male and female
df['Sex'] = le.fit_transform(df['Sex'])

# applying LabelEncoder on multiple class features
df['Deck'] = le.fit_transform(df['Deck'])

df[['Sex','Deck']].sample(6)

```

**Note that in this case the result is a DataFrame as was the input.**

<figure>
<img src="https://content.communere.com/assets/c462b914-d9c0-4db0-b9c7-f7d89a5c306b.JPG" />

 <figcaption style="text-align: center;"> </figcaption>
</figure>
 
 To demonstrate how to deal with other shapes of input, I applied LabelEncoder directly to a DataFrame containing 2 columns. The following code shows different stage results for clarity.

```
df_categorical_data = df[['Sex', 'Deck']] # extracting just 2 of categorical features from the main dataframe
#Convert the dataframe into an array-like of (n_shape,) utilising raval()
df_categorical_data_arryLike = np.ravel(df_categorical_data)
print('The output is an array of shape {} as follows \n {}'.format(df_categorical_data_arryLike.shape, df_categorical_data_arryLike))

#apply fit_transform on the result above
eEncoded_Sex_Deck = le.fit_transform(df_categorical_data_arryLike)

print('The output is an array of shape {} as follows \n {}'.format(eEncoded_Sex_Deck.shape, eEncoded_Sex_Deck))
print(eEncoded_Sex_Deck.shape)

#we can reshape the result utilising .reshape so, the result make better sences
leEncoded_Sex_Deck.reshape(1309,2)
```

<figure>
<img src="https://content.communere.com/assets/e58575e0-6ed5-4555-a61c-bb6f9f4f57ae.JPG" />

 <figcaption style="text-align: center;"></figcaption>
</figure>

Using LabelEncoder for multi-class features may not be the right choice as it may unintentionally bring the notion of ordering among values. For example, in our case, Deck F and B have been encoded by 5 and 1 respectively; However, we can not say F is higher than B. This may result into poor performance in the proposed machine learning model.

### Get_dummies

Provided by **Panda** library, is the most straightforward encoding model. It is intelligent enough to recognise all the categorical features at once in the DataFrame and provide a one-hot scheme code for the categorical data. However, you may not want to encode everything as `get_dummies` does!
To demoneste, I've extracted the features that I was willing to encode them.

```
import pandas as pd
df_categorical_data = df[['Sex', 'Deck','Embarked']]
pd.get_dummies(df_categorical_data)

```

<figure>
<img src="https://content.communere.com/assets/57529686-f6d2-45de-b4f1-f0c870ee69e1.JPG" />

 <figcaption style="text-align: center;"></figcaption>
</figure>

### DictVectorizer

Provided by **Scikit-learn** library, this class of transformer can provide binary one-hot code for non-numerical (string) labels; For the numerical labels, it is required to follow the encoding by OneHotEncoding model to complete the encoding process.
The input of this transformer must be a dictionary-like (feature-value) objects, and output will be an array or a scipy.sparse metrix.

Following our example, first, we need to convert the DataFarme to a dictionary-like object as follows. Note that DictVectorizer is intelligent enough to recognise categorical features in the DataFrame. However, for the sake of presentation, I've extracted a couple of categorical features from the main DataFarme.

```
# convert datafarme to dictionary
df_categorical_data = df[['Sex','Embarked']]
df_categorical_data_dic = df_categorical_data.to_dict(orient='records') # turn each row as key-value pairs
# print
df_categorical_data_dic
```

<figure>
<img src="https://content.communere.com/assets/bfe7c021-2157-4dbc-aa53-337b6dd0a68e.JPG" />

 <figcaption style="text-align: center;"></figcaption>
</figure>

Importing the required library, we can initiate a DictVectorizer object and apply it's fit_transform on the dictionary.

```
from sklearn.feature_extraction import DictVectorizer
# instantiate a Dictvectorizer object
dv_df_categorical_data = DictVectorizer(sparse=False)
# apply Dictvectorizer on the Dictionary
df_categorical_data_encoded = dv_df_categorical_data.fit_transform(df_categorical_data_dic)
print (df_categorical_data_encoded)
```

<figure>
<img src="https://content.communere.com/assets/96a43204-c6bd-4996-92ae-8d289f0d3940.JPG" />

 <figcaption style="text-align: center;"></figcaption>
</figure>

As you can see, the output is an array. To interpret the output, it is beneficial to indicate which column represents which feature by utilising `vocabulary_` attribute as follow:

```
# vocabulary
vocab = dv_df_categorical_data.vocabulary_
# show vocab
vocab
```

<figure>
<img src="https://content.communere.com/assets/afc3d947-9065-4e45-907d-fddc1110c998.JPG" />

 <figcaption style="text-align: center;"></figcaption>
</figure>

### OrdinalEncoder

Provided by **Scikit-learn** library, this transformer provides a single column of integers (0 to n_categories - 1) per each feature. Note that this transformer can not distinguish between categorical and non-categorical features. So, it is beneficial to extract the categorical features that you want to encode before starting the encoding process.
The input to this transformer should be an array-like of the categorical features with integers or strings values. The output will be an array. Note that estimators may interpret the order among values as an actual order among the categorical feature values which is not often the case and leed to poor performance or estimation.

```
df_categorical_data = df[['Sex','Deck','Embarked']]
from sklearn.preprocessing import OrdinalEncoder

# instantiate a LabelEncoder object
enc = OrdinalEncoder()

##apply fit_transform on the categorical featurs
enc.fit_transform(df_categorical_data)
```

<figure>
<img src="https://content.communere.com/assets/31cf22ae-7454-4e5d-b661-d88ac6cc2f9f.JPG" />

 <figcaption style="text-align: center;"></figcaption>
</figure>

To retrive encoded features, we can use `categories_` attribut. Also, it is possible to inverse the transformation to retrive the original values as follow:

```
enc.categories_
enc.inverse_transform(enc.fit_transform(df_categorical_data))

```

### FeatureHashing

Provided by **Scikit-learn** library, The class FeatureHasher utilises " Feature hashing" or "Hashing trick" and provides a high-speed and memory-efficient encoding approach. So, it can be an alternative to DictVectorizer encoder or CountVectorizer in a large, online learning scenario. By default, the input is a dictionary-like of the categorical features. However, it can accept single strings. Check this [link](https://scikit-learn.org/stable/modules/feature_extraction.html) for more information. The output always will be a scipy.spars matrix which we should define the number of its columns when initiating the encoder object. The number of columns should not be too small or large as it may cause low performance or collision in the result.

Following our example, we can apply FeatureHashing as follow. Note that we need to convert our categorical data into a Dictionary.

```
# convert datafarme to dictionary
df_categorical_data = df[['Deck','Sex']]
df_categorical_data_dic = df_categorical_data.to_dict(orient='records') # turn each row as key-value pairs
#df_categorical_data_dic


from sklearn.feature_extraction import FeatureHasher
# instantiate a FeatureHasher object with the number of output columns
h = FeatureHasher(n_features=10)
#apply fit_transform on the categorical featurs dictionary
f = h.fit_transform(df_categorical_data_dic)
f.toarray()[:20]

```

<figure>
<img src="https://content.communere.com/assets/f0b03f0e-487f-4d09-ab36-9dbe353b4abe.JPG" />
 <figcaption style="text-align: center;"></figcaption>
</figure>

In contrast with "OrdinalEncoder", FeatureHashing does not have inverse_transform attribute as it does not remember what the input feature looked like. That is the price paid for its high speed and low memory usage.

## Conclusion

As demonstrated, there are a variety of standard and typical encoding schemes where you can decide based on their required input or provided output type, speed and memory efficiency, their simplicity and more importantly your categorical data essence. It is also possible to combine different schemes or come up with your idea to encode the categorical data. You can check this [Link](https://www.kaggle.com/shahules/an-overview-of-encoding-techniques) for more idea about encoding the categorical features.

Dealing with categorical features

<figure>
<img src="https://content.communere.com/assets/11fb3de3-db47-4ece-8df6-f2936cd6fdd0.jpg" />
 <figcaption style="text-align: center;">source: unsplash</figcaption>
</figure>

When it comes to choosing a Software Development as a Service company for your business, there are a number of choices available for Development and your products or services.

## Why choose remote Software Development as a Service (SDaaS)?

- It gives you the opportunity to tailor your technical projects with a London based programme Manager, allowing for changes to happen in real time
- It allows you to have a project being worked on full time, on a remote basis with our team taking care of pension, salary and the conversations with the development team
- Low carbon footprint from the workforce being fully remote; environmental factors of the office commute will be coming into question over the coming years. Communere is ahead of the curve.
- UK timetable for the development team
- Remote teams are more cost effective, time efficient and agile

Communere is Londons leading remote Software Development as a Service team.

Our product is our people; we have a remote network of world class software developers who specialise in Angular, Artificial Intelligence, Backend, Frontend, Fullstack, React, React Native, GraphQL, among others, who are trained professionals who receive training both in the office and online to ensure they are knowledgeable and skilled to build your products.

The team at Communere have worked on projects for small and scaling organisations, they have a can-do approach with the ability to pick up a project and deliver. We work in an agile methodology which means our customers have control and insight into how their project is progressing. Our team has some of the leading technical skills required for organisations over the coming years and we would love to help you achieve your business goals.

Our passion for providing exceptional customer service, technical advisory and delivery, and satisfaction is embedded in our core values and is at the centre of everything we do.

Communere; why we are the best from the rest

Last summer, my family and I went to Paris. One night. We were out for dinner and were looking for a parking space near the restaurant we had targeted. While we were hunting a parking space near the restaurant, we got distance from it. So, we had to walk down a couple of streets to get to the restaurant. As we start walking, we felt unsafe since many strangers were wandering around or sitting in a group in the dark. We were worried about how safe it will be to return to our car in a couple of hours. We decided to go back to our car and find somewhere more close to the restaurant. Sadly, we couldn't find any parking space. So, We decided to call the restaurant and ask for their local expertise. We wondered **how safe this neighbourhood is**?

The restaurant manager tried to convince us that _the neighbourhood is very safe due to the existence of many luxury pubs and restaurants in it_.

Having this stressful experience during my holiday, I decided to examine the hypothesis that the restaurant's manager put forward to us. **I want to test whether a neighbourhood with a high number of venues are safe or not in comparison with the area with less number of venues?**

## Case study, Tools and Data selection

I've selected Surrey County in the UK as a case study and utilised Jupyer on IBM Developer Skills Network Labs to implement my model. The information required to conduct this analysis is geographical coordinates, criminal records and available venues in each specific part of the selected region.

The dataset, including information related to the crime committed in each ward in this county, are available on [Surrey-i website](https://www.surreyi.gov.uk/) via this [link](https://www.surreyi.gov.uk/dataset?q=Master Crime Category by Ward ). Although the dataset includes Geocode of each ward, due to license requirement, I was unable to use this information. I couldn't find any free dataset or specific website with required geographical coordinates information where be able to extract the required data systematically. So, I choose a postcode in the centre of each ward manually. Then, I merge this information with the dataset containing all available postcodes in the UK and their latitude and longitude coordinates where I can estimate the latitude and longitude of the centre of each ward.

All the datasets that I've used for this study are available in the following links:

- [Master Crime Category by Ward.xls](https://www.surreyi.gov.uk/dataset?q=Master Crime Category by Ward https://www.surreyi.gov.uk/dataset?q=Master Crime Category by Ward )

- [Wards by postcode.xls](https://github.com/SepidehN/Coursera_Capstone/blob/master/Wards_by_postcode.xls)

- [UkPostcodes.csv](https://www.freemaptools.com/download-uk-postcode-lat-lng.htm)

Other required information is the venues' data associated with each specified coordinates in each ward. To do so, I have utilised **Foursquare API**. Since I am interested in how busy/crowded the areas are, I calculated the total number of venues in each specified location and merged this information to the primary data. Now, I am ready to examine and analyse the wards based on their popularity and criminal records. The details of data acquisition, data cleaning and preparation can be viewed on this [Jupyter notebook](https://github.com/SepidehN/Coursera_Capstone/blob/master/TheBattleofNeighborhoods-Part2.ipynb)

## Data Analysis

To do this, I've used the _k_-mean clustering algorithm, a simple but powerful unsupervised machine learning technique. K-mean learns from the properties of data and divides it into different (hopefully optimal) groups. Many clustering algorithms are available. But perhaps the most widely used one is _k_-means available in sklearn.cluster.KMeans.

One of the main drawbacks of the _k_-mean algorithm is the difficulty of predicting the number of suitable clusters. To overcome this, I ran the algorithm with different parameter _k_ and analysed the outcome of each cluster and finally, I have chosen the best _k_ that can divide the wards more explicitly.

After applying the algorithm and regulating the parameters in the algorithm (i.e., k) and Foursquare API (i.e., limit and radius), I have discovered that it is sensible to cluster Surrey's wards into 5 different clusters as presented and visualised in following table and image respectively.

| Cluster | color | Label |
| :-----: | :----: | :--------------------------------: |
| 0 | Green | Moderate density & Moderate risk |
| 1 | Blue | Very high density & Very high risk |
| 2 | Purple | Moderate density & Moderate risk |
| 3 | Red | Very low density & Very low risk |
| 4 | Olive | low density & High risk |

<figure>
<img src="https://content.communere.com/assets/a15b16ee-7c69-479b-864f-3cc81ba528e5.JPG" />
 <figcaption style="text-align: center;">Visualisation of the clusters on the map</figcaption>
</figure>

Cluster 1, which includes wards in the centre of two big towns in Surrey area (i.e., Guildford and Epsom), shows that more venues can cause more risk in the area. As the table below shows the number of crime committed in this cluster are significantly high. On the other hand, Cluster 3, is associated with wards that have a very low density of venues and a slight risk in terms of crime committed. However, Cluster 4, despite having a low density of the venues, it has a high risk where make these wards undesirable since they are neither enough accessible nor safe enough. These wards are mostly inside the M25 or near big town such as Woking or Redhill. Cluster 2 and 0 are slightly similar. They are wards with moderate density and relatively moderate risk. However, as presented on the map, Cluster 0 includes wards where most are located inside the M25, which make increases in specific categories of crime such as theft handling and vehicle interference in these areas. In general, the model confirms we should expect crime rises in busy/popular areas. The details of the model can be viewed on this [Jupyter notebook](https://github.com/SepidehN/Coursera_Capstone/blob/master/TheBattleofNeighborhoods-Part2.ipynb)

| Cluster | \# of venues | TNO | Theft handling | vehicle interference |
| :-----: | :----------: | :--: | :------------: | :------------------: |
| 0 | 42 | 474 | 114 | 13 |
| 1 | 78 | 1860 | 723 | 6 |
| 2 | 39 | 474 | 93 | 4 |
| 3 | 28 | 207 | 35 | 3 |
| 4 | 34 | 983 | 231 | 8 |

## Conclusions

Utilising clustering technique, I was able to group different wards in unique clusters based on the number of venues and crime features in those areas. I built the k-mean clustering model and regulated different parameters in the model, such as the number of clusters and identified 5 clusters. This model indicates that we can not assume that areas with a high number of venues (popular places) are safer than areas with less number of venues.

Battle of Wards

This article will present the process of configuring **Azure DevOps** Services to an existing project. Our development flow is somehow similar to the **Git-Flow** in the way that we have two main branches `master` and `develop` and we’re going to to trigger CI/CD operations on new commits that are pushed to these branches. For this purpose, we are going to use Azure Pipelines.
**Azure Pipelines** is free for public projects, but if you use private projects, you can run up to 1,800 minutes (30 hours) of pipeline jobs for free every month. Learn more about how the pricing works based on [parallel jobs](https://docs.microsoft.com/en-us/azure/devops/pipelines/licensing/concurrent-jobs?view=azure-devops).

 
# Creating a new Pipeline
To start, we create a new Pipeline using `Create Pipeline` button and choosing the appropriate repository on an existing project.

![](/images/s_949952C31EB49A2F50BAFD1B9B1F99BB29463D8B034D14BDC591001FD197B2B7_1569904531860_image.png)

On the next page it shows a couple of Pipeline templates to use from.

![](/images/s_5F6D7052D0C8FF54DBC0E5630BF834E26B49652948EFE8506C966D6E20CFF65B_1572428345641_image.png)

If you’re creating a whole new pipeline from scratch these templates are a good starting point. If you already have a YAML file then you can chose the `Existing Azure Pipelines YAML file` and on the next page choose the appropriate YAML file.

After choosing the proper pipeline template (for example .NET Desktop), we are presented with the YAML preview page in which we can modify the generated `.yml` file. Here we can see actual steps of our Pipeline. A **step** is the smallest building block of a pipeline. For example, a pipeline might consist of build and test steps. A step can either be a script or a task. A **task** is simply a pre-created script offered as a convenience to you.

![](/images/s_949952C31EB49A2F50BAFD1B9B1F99BB29463D8B034D14BDC591001FD197B2B7_1569905683450_image.png)

**YAML** is a storage human friendly serialization standard (like JSON) which basically contains key value pairs. Here we are using it to describe different properties of the Pipeline. The first property is the `trigger` property. A **trigger** is something that's set up to tell the pipeline when to run. You can configure a pipeline to run upon a push to a repository, PR creation, at scheduled times, or upon the completion of another build. Note that YAML PR triggers are currently only supported in GitHub and Bitbucket Cloud. If you are using Azure Repos Git, you can configure a [branch policy for build validation](https://docs.microsoft.com/en-us/azure/devops/repos/git/branch-policies?view=azure-devops#build-validation) in order to trigger your build pipeline for validation. As it can be seen the default trigger is set to build upon push to the `master` branch.

Next, there is some configuration like `vmImage` to be used for build and some variable declarations, followed by the actual steps for our Pipeline in the `steps` definition.
A couple of steps are per-defined for us by the template. In this example the steps are:
`NuGetToolInstaller` : Fetches the proper NuGet tools required by the next step.
`NuGetCommand` : Installs the dependency packages.
`VSBuild` : Builds the solution.
`VSTest` : Runs the tests.

Each task is followed by a number which defines the version of that task. For example `VSTest@2` means version 2 of the `VSTest` task.
Also the YAML editor adds a little button called `settings` which we can use to modify the tasks visually.

If you’re happy with the configs you can click `Save and run` which will create the file in a new commit and will ask for the commit message and the destination branch, and run the Pipeline.

![](/images/s_949952C31EB49A2F50BAFD1B9B1F99BB29463D8B034D14BDC591001FD197B2B7_1569908770607_image.png)

If everything is configured correctly the Pipeline will successfully run to completion and you’ll get the delightful green tick mark.

![](/images/s_949952C31EB49A2F50BAFD1B9B1F99BB29463D8B034D14BDC591001FD197B2B7_1569909304410_image.png)

![](/images/s_949952C31EB49A2F50BAFD1B9B1F99BB29463D8B034D14BDC591001FD197B2B7_1569909333976_image.png)

 
# Packaging into NuGet artifacts

And that it for most basic use cases. But in some cases you might want to create a NuGet package and push it to your NuGet repository. In that case you neeed to two additional steps:

- **Fetching** some NuGet packages from custom NuGet repository
- **Packing and pushing** the compiled assemblies to the custom NuGet repository.

 
## Creating NuGet feeds
To create a new NuGet feed, you can go to the **Azure Artifacts** and click on `Create Feed`.

![](/images/s_949952C31EB49A2F50BAFD1B9B1F99BB29463D8B034D14BDC591001FD197B2B7_1569918237021_image.png)

Then fill in name and permissions and `Create`.

![](/images/s_949952C31EB49A2F50BAFD1B9B1F99BB29463D8B034D14BDC591001FD197B2B7_1569918071622_image.png)

It is also possible to add some other custom sources to this by going to feed settings and choosing the **Upstream sources** tab. We can see there are a couple of sources defined already.

![](/images/s_949952C31EB49A2F50BAFD1B9B1F99BB29463D8B034D14BDC591001FD197B2B7_1569918660083_image.png)

Now if you head back to the Pipeline editor, you should be able to choose this feed as NuGet source.

![](/images/s_949952C31EB49A2F50BAFD1B9B1F99BB29463D8B034D14BDC591001FD197B2B7_1569918392444_image.png)

 
## Packing NuGet packages and push them to the artifact repository
For the second task we have to add two new NuGet tasks for packing and pushing the final package to the NuGet feed.

![](/images/s_949952C31EB49A2F50BAFD1B9B1F99BB29463D8B034D14BDC591001FD197B2B7_1569919647126_image.png)

Note that we must configure the versioning based on the company’s policies. For example, here the following versioning policy has been selected which uses date and time policy

![](/images/s_949952C31EB49A2F50BAFD1B9B1F99BB29463D8B034D14BDC591001FD197B2B7_1569937167532_image.png)

Note that if you already have set up the automatic versioning at project level (on the `.csproj` ) you can set the Automatic package versioning here to `Off`.
After a successful pipeline run, the resulting package can be found at the artifacts section.

![](/images/s_949952C31EB49A2F50BAFD1B9B1F99BB29463D8B034D14BDC591001FD197B2B7_1569937714792_image.png)

Here the `TestLibrary` is the artifact that was created by our NuGet task. You can see that there are two other NuGet packages inside our artifact’s feed and that’s because by default when are are creating an artifact feed, it’s going to add a couple of other Upstream feeds like nuget.org to it.

 
# Additional notes
- Instead of using Microsoft’s clouds, you can run these Pipelines on your own server for free using your own build agent.
- Packages with a **pre-release** tag will not be shown on NuGet clients by default until the consumer opts-in with a check-box or command line switch.

 
# References
- [https://docs.microsoft.com/en-us/azure/devops/?view=azure-devops](https://docs.microsoft.com/en-us/azure/devops/?view=azure-devops)
- [https://docs.microsoft.com/en-us/azure/devops/artifacts/concepts/best-practices?view=azure-devops](https://docs.microsoft.com/en-us/azure/devops/artifacts/concepts/best-practices?view=azure-devops)
- [https://docs.microsoft.com/en-us/nuget/create-packages/prerelease-packages](https://docs.microsoft.com/en-us/nuget/create-packages/prerelease-packages)
- [https://www.visualstudiogeeks.com/azure/devops/perfecting-continuous-delivery-of-nuget-packages-for-azure-artifacts](https://www.visualstudiogeeks.com/azure/devops/perfecting-continuous-delivery-of-nuget-packages-for-azure-artifacts)

Getting started with Azure DevOps Services

![](https://content.communere.com/assets/2966f4c5-b916-4d94-bd54-6103efa222b5)

## Before start

[Online Multiple Traveling Salesman Problem Solver - MTSP Solver](https://ralthor.github.io/mtsp)

[Source code for Multiple Traveling Salesman Problem Solver](https://github.com/ralthor/GeneticAlgorithm-TSP/tree/feature-multi-tsp)

<hr/> 

## What is Traveling Salesman Problem?

Assume you are a traveling salesman, selling coffee. You want to start from your home, sell coffee to costumers at different places and return to your home. You are intelligent and you want to minimize the distance you travel. The route you choose will be a solution to the traveling salesman problem.

## How to solve?

Traveling Salesman Problem (TSP) is proven to be unsolvable in polynomial time, therefore lots of heuristic/meta-heuristic algorithms are used to solve that and give near optimal solutions. [This link](https://github.com/parano/GeneticAlgorithm-Solving-TSP) is parano's genetic TSP solution. It is also [available online](http://parano.github.io/GeneticAlgorithm-TSP/).

### Genetic Algorithm

If you are going to read the code mentioned above, it is good to know Genetic Algorithm and search for its methods inside the code.

![](https://content.communere.com/assets/dbbfd5c1-310d-479a-8ffb-b2b0bde1192c)

1. Population: in genetic algorithm, there is a pool of solutions, initialized randomly. Each solutions is called an individual, and the solution is coded inside it. In TSP, each individual is a rotating array of numbers, which shows a route between costumers. For example, `3 0 4 2 5 1` means going from `3` to `0`, then `4`, then `2` and so on. Note that it also goes from `1` to `3`, because the list is rotating. Therefore, it is equivalent with for example `2 5 1 3 0 4`.
1. Initialization: random evaluation of the population is done here.
1. Fitness function: since it is important to compare different solutions, a fitness functoin is defined. It accepts an individual and returns how much it fits. For TSP, it is defined as `evaluation` method, which returns the sum of the distances between the costumers. However, since it returns smaller values for better solutions, the fitness is defined as `1/evaluation`, which returns higher values as the given solution (passed individual) gets better.
1. Crossover: combines two individuals and creates new one(s). It is one of the most important parts of a Genetic Algorithm solution, and there are several scientific papres proposing different crossovers for TSP. The one that is used in parano's code, is a simple greedy algorithm which creates two children for each pair of parents. It starts from a random starting node (costumer), and add it to the child. Then it checks the next nodes after the starting node in both parents. Next node of the child will be selected from the parent in which the next node is closer to the starting node. For example, if starting node is `7`, and the next nodes in parents are `3` and `9` respectively, if `distance(7, 3) < distance(7, 9)` then `3` will be selected, otherwise `9` will be the next node. The other child is found going to the previous node of the selected node, instead of going to the next.
1. Natural selection: or simply parent selection, selects a pair of fit parents from the population. They will be sent to crossover to create the new generation.
1. Mutation: to search all of the solution space, the mutation is also happen to the population. It is done with low random probabilities such as 0.01. For TSP, again several different mutations are introduced. For example, swapping two random indices in the array, or reversing a random subset of the array in place. It is shown in the above image.
1. Solve method: almost all of the previous methods are called in this method. It means that for a number of generations, we create new populations by using crossover and mutaion over best fit individuals. A semi-code for this method is as below:

{% highlight javascript %}
function solve() {
let population = initialization();
for(let i = 0; i < generationCount; i++) {
let newPopulation = [];
for(let j = 0; j < population.size(); j+= 2) {
let parent1 = selectFitParent(population);
let parent2 = selectFitParent(population.filter(p => p != parent1));
let children = crossOver(parent1, parent2);
children[0] = mutate(children[0]);
children[1] = mutate(children[1]);
newPopulation.push(children[0]);
newPopulation.push(children[1]);
}
population = newPopulation;
}
}
{% endhighlight %}

In the code repository, the `for` loop is replaced with a method called `GANextGeneration`, which is called using Javascript `setInterval` method.

## More than one salesman...

Assume you are living with your friends or family, and they decide to help you. You all start from home, each travels to a different set of costumers and returns. The new problem you are facing is called Multiple Traveling Salesman Problem (MTSP).

It can be divided into two problems:

1. Select destinations for each salesman,
2. Solve TSP for each salesman.

However, there are some ways to convert MTSP to TSP, and apply the previously built solutions to the new problem.

## Multiple salesmen? It is easy!

I changed parano's solution, to work for Multiple TSP (MTSP), in which there are more than one traveling salesman. It is an [online multiple traveling salesman problem solver](/mtsp). The [source code](https://github.com/ralthor/GeneticAlgorithm-TSP/tree/feature-multi-tsp) is also available.

<a href="/mtsp"><img src="https://content.communere.com/assets/images/threeSalesmen.png" /></a>

Two easy steps I took to change the base code from TSP to MTSP solver:

### 1. Added source nodes for each salesman

For TSP, we have only one source node. I added several other source nodes at the same place with the main source.
This way, one of the possible solutions for 10 nodes and two salesmen may be `3 8 4 0 5 7 1 9 0 6 2`. Since the list is rotating, it can be read as `5 7 1 9 0 6 2 3 8 4 0`. Different salesmen's paths are separated with 0, which is the index of the source node.

However, we may end up paths with length 0, for example `1 7 4 8 0 0 3 5 2 9 6`, in which one of the salesmen started from 0 and immediately went to 0.
To avoid this, the distance between source nodes is assumed to be infinity.

### 2. Changed the evaluation method to consider multiple salesman

Previous evaluation method was returning the sum of the distances. In other words, it was calculating the total distance the salesman travelled. In the code, it is implemented like this (of course after a `for` loop calculating it):
```js
return sum;
```

In order to maintain all of the salesmen's travelled distances almost the same, I changed above code to this:

```js
var L2 = sumForSalesman.reduce((t, x) => t + x\*x, 0);
return L2 / sum;
```

In above code, `sumForSalesman` is an array of the distance each salesman traveled. I square each of its members and calculate the sum of them. If there is only one salesman, the result will be the same as previous implementation. Therefore nothing is changed from the TSP solver perspective. However, when there are more than one salesman, above formula returns larger values for more unfair routes. For example, if we have two salesmen and they traveled 10 KM in total, the result will be different for this cases:

- both traveled 4 KM: `(4*4 + 4*4) / 8` which gives `4`
- one traveled 5 KM and the other 3 KM: `(5*5 + 3*3) / 8` which gives `4.25`
- one traveled 1 KM and the other 7 KM: `(1*1 + 7*7) / 8` which gives `6.25`

If you think this policy won't give the opimum solution every time, you are right! Here is an example, for the same path, 2+6 and 5+5 solutions are prefered over 1+6, although the sum for the last is smaller than those two.

## Conclusion

There are lots of interesting problems to solve with regards Traveling Salesman Problem. For example, consider the salesmen are not bound to start and end from the same location, or some of them are constrained to visit a certain number of nodes (costumers), or travel a certain distance at most. In addition, several other crossover and mutation methods may be tested to help the method to find an acceptable answer earlier.

Implementing a Solution to Multiple Traveling Salesman Problem

![](https://content.communere.com/assets/2762cb67-9641-4321-a483-3ec60dde4eab)

Are you trying to develop a simple AI agent to win a game? Maybe Q-Learning is a right choice for you. Applying Q-Learning, your system will learn based on its previous experiences. It means in order to choose between its possible actions, it should have experienced a similar situation. Base on the environment, that may end up the situation when we leave two (or more?!) of our agents to play against each other and learn (Google DeepMind!), or just put the agent in the world to learn from its experiences!

## Q-Learning story

The learning process in Q-Learning is based on scoring the next possible actions. These are called Q-values. It means looking at the environment during the learning process, the agent scores each of its actions based on its experience. For example, in a maze game, such as PacMan, if some actions lead the agent to win the game, the scores of those actions are boosted to increase their chance to be selected later. On the other hand, if some actions lead to fail, for example running into a ghost, the scores are decreased to avoid selecting them in the future.

## How to set up the Q-Learning process?

![](https://content.communere.com/assets/de25a3a2-a243-475e-bf72-974a120ebc31)

The learning environment includes several world states. The agent moves from one state of the world to another by selecting an action. The states can be represented by tables, each contains several cells. For example, in a PacMan game, a ghost stands at a certain cell. For each cell, adjacent cells are available for the agent to choose for its next action. This is where the scores (Q-values) are required. Therefore, there is a score for each next action, say the next cell. These scores are set to a fix value, say zero, at the beginning, because the agent has yet no idea of the outcome of its actions. However, there is a reward/punishment system in Q-Learning method, which effects the scores during the learning process. These rewards and punishments are certain cells, with known scores. For example, the score for the exit cells and ghost cells can be +1 and -1 respectively. The run between the start of the game and when it is finished (or we consider it enough for learning) is called an _episode_. A process of Q-Learning may consist of thousands of episodes or more.

The definitions presented in this step are:

- state of the world (state): a representation of what happen in the world, where the agent is able to see the situation and select its next action. For the agent, the state may be desirable, undesirable or something between these two.
- actions of the agent: actions change the status of the world, and may lead to a desirable/undesirable state. These must be selected _intelligently_, and the learning process makes it possible.
- score or Q-values: The intelligence behind Q-Learning is based on the score it assigns to the pairs of states and actions. In other words, when the agent is in a _state_, and it selects an _action_, the agent wants to know how much score it will gain. This makes it possible for the agent to select the best action for the current state of the world, even if no direct reward/punishment is given at the moment.
- reward and punishment: although at the beginning there is no meaningful score for the most of the world states, if the agent reaches some world states, it is rewarded or punished. Those states are not necessarily the end of the game.
- episode: The run between the start of the game and when it is finished, or when it is decided to apply the scores.
- learning: ending the game (either by loosing or winning, or even falling into a loop which may interpreted as a fail) gives the opportunity to see which states and actions lead the agent to reach the current state. Q-Learning changes the Q-values for all of the path. However, this is a weighted change, which is dependent to both the times the state, action pair is updated and the distance of it to the end of the episode. For example, the tenth update has more effect than the fifteenth, and the adjacent decision to the end of an episode is more effected than the first decision the agent made in that episode.
- state, action pair: it is shown by `(s, a)`. In Q-Learning, a score is assigned to each pair, which is written `Q(s, a)` and read Q-values of performing action `a` in state `s`.

## Exploration vs Exploitation

![](https://content.communere.com/assets/7aa4ca20-fe49-4e2f-bedf-1b065a90691a)

Put the agent at the start cell, and let it takes its first action! At the beginning, because of the unknown scores, its actions are chosen randomly. However, when it reaches its first win/lose cell (reward or punishment cell), the cells in the path will be updated. In order not to take similar actions on the first positive response, or avoid a right path if it is failed somewhere at the end, an exploration mechanism is embedded into the learning phase of the Q-Learning. It means during the game, there is a probability that the agent selects an action except the current best action. This is called the exploration probability. This is decreased during the learning process. It helps the agent to explore the environment at the early stages of the learning process, and to focus on improving its final steps at the end.

## Lets talk formula!

![](https://content.communere.com/assets/be03dd7b-373f-4f9f-9ccc-0e0b00f8bec8)

Before coding, let's look at the formulas. They help to implement what everyone knows as Q-Learning. This is the learning formula, based on [Wikipedia](https://en.wikipedia.org/wiki/Q-learning):

$$Q^{new}(s_t, a_t) \leftarrow (1-\alpha).Q(s_t, a_t)+\alpha . (r_t + \gamma . max_aQ(s_{t+1}, a))$$

In this calculation:

- $$Q^{new}(s_t, a_t)$$ is the new Q-value (score) for this state, action pair
- $$\alpha$$ is the learning rate, between 0 and 1 (0 means no learning and 1 means dismiss old learnings)
- $$\gamma$$ is the discount value. This is between 0 and 1, and shows how much the score is important. For 0, even final state, actions are considered worthless. For something in between, the calculated score becomes less important as it goes backward from the final score, say end of an episode.
- $$r_t$$ is the reward. It may be zero if the agent has not reached the required state. However, if it has reached, it shall be rewarded!
- $$max_aQ(s_{t+1}, a)$$ is the Q-value for best available action in state _s_.

## Developing the Q-Learning environment

Here is a simplified Javascript code for one episode of the Q-Learning process.

```js
let pairs = [];

let s = environment.start();
let a = null;
for (let i = 0; i < maxEpisodeLength && !environment.isFinal(s); i++) {
let actions = environment.actions(s);
if (epsilon > Math.random()) // exploration vs exploitation
a = actions.randomElement(); // explore
else
a = selectBest(s, actions); // exploit

 let candidateState = environment.applyAction(s, a)
 if (pairs.map(p => p[0]).findIndex(state => state == candidateState) != -1)
 	continue; // if the state is repeated, it is not selected.

 pairs.push([s, a]);
 s = candidateState;

}
let score = 0;
let reward = 1;
if (environment.isFinal(s)) {
score = environment.isFinal(s);
if (score > 0)
reward = 1.1;
}
else
score = bestScore(s);

for (let i = pairs.length - 1; i >= 0; i--) {
let prevScore = pairs[i] in qvalues ? qvalues[pairs[i]] : 0; // retrieve the previous score of this pair, if any.
qvalues[pairs[i]] = ((1 - alpha) _ prevScore + alpha _ (reward _ discount _ score));
score = qvalues[pairs[i]];
}
```

The complete working code is available in [this github repository](https://github.com/ralthor/q-learning-javascript).

### Implementation notes:

- To develop Q-Learning, I make a big loop over all of the episodes I want to have, which could be unlimited.
- Inside that loop, I put the agent in the starting state, and let it react to the environment based on its Q-values.
- When the agent reaches the end of the current episode (either by reaching a final state or the end of its permitted action numbers), I have made a list out of its state, action pairs.
- To avoid loops inside the list of state, action pairs, the actions which create a loop is not selected. It means that the agent waits on that cell to select another action on its next chance.
- Starting from the end of the list, I apply the learning formula to all of the state, action pairs it traversed.
- Instead of adding the reward, I simply multiplied it. It helps not to be aware of reducing it while going backward through the list.

## What to search/read next?

These topics maybe interesting:

- Q-Learning variations and alternatives
- Different policies for learning
- Deep Q-Learning

## Next post

In the next post, I will implement a Q-Learning powered agent playing a game. I like it to be snake, since I have created the environment [here](https://github.com/ralthor/Snake). However, as the environment gets larger, storing all of the state, action pairs gets harder. Therefore, the next implementation may use Deep Q-Learning!

A brief description on Q-Learning and how it is implemented

Have you ever attempted to set up an automated backup of your MySQL database? The process of creating a scheduled backup is not as easy as SQL Server. In MySQL, we must use the command line to make this process automatically. The method uses a batch file to automatically backup and zip files each night at midnight. This happens using `mysqldump` command-line tool using Windows Task Scheduler:

![](https://content.communere.com/assets/80b05040-a71f-49de-97ea-880ccba2075b)

Here’s the command which is being called behind the scene:

```bash

set dbUser="DB_USER"
set dbPassword="YOUR PASSWORD"
set backupDir="PATH"
set dbName="DB_NAME"
set mysqldump="C:\Program Files\MySQL\MySQL Server 8.0\bin\mysqldump.exe"
set zip="C:\Program Files\7-Zip\7z.exe"

%mysqldump% --host="MYSQL_SERVER_ADDRESS" --user=%dbUser% -p@%dbPassword% --single-transaction --add-drop-table --databases %dbName% > %backupDir%\%dirName%\output.sql
%zip% a -tgzip %backupDir%\%dirName%\%fileSuffix%output.sql.gz %backupDir%\%dirName%\output.sql
del %backupDir%\%dirName%\output.sql
```

This command, first backups the database then uses `7zip` to zip the backup. Then the `sql` is deleted as we don't need it anymore.

![](https://content.communere.com/assets/7994d38f-17e9-497e-be41-af0698179224)

Communere’s Blog