Reducing Customer Churn In The Telecommunication Industry By Use Of Predictive Analytics

Abstract

Customer churn is a big problem in various businesses and especially so in the telecommunication industry. When a business loses its customers, it loses the revenue that was being generated from the customers and possibly revenue from potential customers who receive negative marketing from customers who churn. Managing customer churn in the Kenyan telecommunication industry has been largely ineffective due to the reactive approach where by churn is just a metric that is reported by the business after a certain period. The objective of this study is to show how we can use predictive analytics to proactively identify customers who are about to churn. By doing so businesses can take measures to prevent or reduce churn and therefore increase their customer retention. This was done by identifying features that are most important in predicting churn, developing, implementing and testing a churn prediction models and evaluating the performance of the models. While there exist different approaches to solving the churn problem, machine learning was used to do the churn prediction based on various customer attributes such as age, usage, gender, etc. Since there exists multiple algorithms to do this kind of machine learning, this research implements four of them and does a comparison to see which one would be the most suited based on their performance. The final result shows which features can be used for churn prediction which were Registration Document, Age on Network, Subscriber age and Talk Time. The importance of each of these features was also shown. Classification algorithms that were used are Random Forest, K Nearest Neighbors, Naïve Bayes and Neural Network. The end results show how the algorithms perform in terms of accuracy and execution time.