Customer Churn Prediction For Business Intelligence Using Machine Learning

The telecom industry is characterized by intense competition among industry players on every scale such that customer churn prediction and management is, by far, one of the highly ranked challenges faced by these organizations. There are, however, a variety of machine learning techniques utilized to predict a customer who will likely churn from a telecom firm to another. This paper sorts to solve a classification and prediction problem in which customers who are likely to churn and those who will not were supposed to be predicted from the Teldata data set. To achieve this, SVM (liner, RBF, polynomial, and sigmoid kernels), MLP (with Adam, SGD, and LBFGS algorithms) and Neural Networks (with Adam optimization technique) machine learning algorithms were employed and results compared to choose which technique best fits the problem. Results showed that Neural Network with Adam optimization technique outperformed the other techniques listed.