Recommendation Systems in Telecom Industry

Recommendations for Plans

Congratulations! if you have stumbled across this usecase,it means you are on the right path in your journey.

We all are surrounded by recommender systems all around us,whether it be a movie or a tv series we are watching on Netflix or Amazon Prime or searching a particular product on a ecommerce website.Even the text you type in your keybad while texting uses some sort of recommendations ( to be specific it uses a concept called as federated learning)

In the current period of data over-burden, it is getting progressively difficult for individuals to discover applicable substance. Recommender frameworks have been acquainted with assistance individuals in recovering possibly valuable data from a colossal arrangement of decisions. It is a framework that proposes items, administrations, data to clients dependent on the examination of information and has gotten progressively noteworthy in taking care of the data blast issue.

As of late, huge advances have been taken toward giving customized suggestion administrations to a wide assortment of electronic applications in web based business, e-learning, and e-administration. Be that as it may, this framework is shockingly not present in the telecom part starting today, however it is a need. Thus, through this blog we will see the requirement for recommender frameworks in telcom industry to help the client base in choosing the most fitting telecom plans/items available according to their needs. we will see various types of recommender systems,discuss their pros and cons The blog will close with the turn of events and execution of a customized proposal framework for telecom items/administrations by mining certain rules and insights from information to recommend the correct telcom bundles/plans/items for clients.

Getting started…

Telecom Industry do possess huge amounts of information, maybe more than any organization in the world . In any case, the battle lies in catching the insights from this unstructured datset and drawing business insights using the knowledge of ML .

It is one thing to have meaningful information gathered in a conventional manner, yet it is another to get this information and map it to a business usecase and get your company some huge bucks.Most telcos administrator’s site page today, comprises of an immense number of decisions, which thus turns out to be progressively difficult for a client to discover his/her ideal decision of item.

Wouldn’t it be good if the customers are shown recommendations on different extra items/groups dependent on decisions made by existing clients before?

Usage of Recommender System

Recommendation frameworks can be calibrated or fine tuned to take into account client profiles/inclinations,plans purchased in past to give them choices they might make in the near future The upside of this framework is to recommend the correct things (base plans or extra items) to specific clients, in view of their past behaviour so that company gets a long term satisfied customer,because at the end of day your organization needs to make profits.

Types of Recommender System

Through this blog I will discuss mainly two types of recommender systems.

1 Collaborative filtering

2 Content based filtering

Collaborative Filtering

The key concept in collaborative filtering methods is that they are collaborative, i.e. they leverage other user’s data. If you are trying to guess whether or not you will like a certain movie,or we are figuring out whether a plan is useful or not you might ask people with similar interests what they thought of that movie or You also might ask these people what other movies they liked and gather a list of recommendations. Collaborative filtering methods predict the target user’s rating of a given item using similar users’ ratings of that item. in our case we can predict the the best plan for a customer given past history of other customers.

Your favourite streaming platform Netflix uses a hybrid approach of this type of recommender system!! It also comes in two flavours

  1. Memory-based or neighborhood-based methods are a simple way of harnessing the ratings of others users to predict ratings. They themselves come in two varieties: user-based and item-based. User-based collaborative filtering forms a pool of similar users and averages of their ratings of the target item.In our case it would average out the plan chosen by most users. Item-based collaborative filter forms a pool of similar items, and averages the target user’s ratings of those items. For example

user1={plan1,plan2,plan3}

user2={plan2,plan3}

user3={plan3,plan4,plan2}

New user4 comes and suppose he chooses plan3 so most likely our algorithm should recommend this person plan2 to make you understand in an intuitive manner.

  1. Model-based methods take collaborative filtering a step farther, and uses machine learning and probabilistic models. Dont worry I will not trouble you with these technical jargons. :)

Content-based Recommender Systems

Unlike collaborative filtering methods, content-based recommender systems do not use other user’s ratings at all. Instead, they utilize descriptive keywords associated with each item to make recommendations. This is quite useful, because the only rating history we need to make predictions is that of the target user.

For example we find that plan1 is similar to plan2 using some transformations ,doing some feature engineering(on basis of some common features) Notation we are using..

p1={p2} → p1 is similar to p2 and so on.

p2={p3,p4}

p3={p4,p2}

if a person has purchaseed plan 2 then in near future there is a higher probability that he might choose plan4 just to make you understand in an intutive manner.

Most popular algorithm for such a usecase is Apriori algorithm.

There are four performance metrics within the Apriori algorithm. These parameters are used to exclude association rules in the result that have support and confidence lower than the minimum support, confidence, lift, or length, respectively.

Suppose we have the dataset like this…

This is hypothetical data on basis of this using our algorithm we can get the most frequently used data plans by our customers. For the sake of this example lets assume we get results like these..

I have not discussed the technical details of the algorithm as there are so many articles and blogs over the internet to make you understand and implement this from scratch.The purpose of this blog was to make you aware about one of the usecases that can be deployed in real time for telecom industry.

As far as implementation goes I have deployed and built a movie recommender system ( not something related to Telecom) which is an end to end project. It will give the recommendations of movies and also perform sentiment analysis on basis of reviews of that movie.

Hope you have learnt something new by reading this blog :)

Link https://movie-recommend-with-sentiment.herokuapp.com/

By Kunal Verma

Machine Learning Engineer at Inventum Technologies Pvt Ltd

ML engineer | Data scientist

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store