Telecommunication Services Churn Prediction - Deep Learning Approach

Abstract

Churn is a phenomenon that concerns the majority of companies, especially in the telecommunication industry. This paper describes experiment on data provided by the telecommunications company - Orange, for predicting churn. The preprocessing phase of the experiment included removal of missing values and redundant data, Lasso and manual feature engineering. Convolutional Neural Network was applied as classifier on preprocessed one-dimensional dataset with accuracy of 98.85%. Proposed model can be applicable in telecommunication systems for detection.