Customer Departure Prognostication: Modern Data Science Methods

Churn prediction involves finding loss of customers over a service for a specific period of time. Prediction of churn rate gives a leeway for many companies. In telecommunication sector churn rate is important because it can be used to predict potentially customers who can retained. In this work we delve into machine learning approaches to predict churn rate. We have experimented with five different algorithms which are Logistic regression, KNN, SVM, Random Forest, XG Boost algorithm and lastly neural network model. Among the five ML approaches XG boost gives the highest accuracy of 82% while KNN give the least accuracy of 76%. To further improvise the performance, we have used a neural network model which gives a training accuracy of 92%.