Telecom Churn Prediction Using Python

In this study, a convolutional neural network (CNN) model was put out for the prediction of customer turnover in the telecom sector. In earlier studies, numerous supervised machine learning models were developed and applied to forecast client attrition. However, in order to create these models, attribute selection must be done by humans, which is tedious, time-consuming, tailored to specific datasets, and usually results in attribute selection problems. In order to eliminate human attribute selection and accompanying issues, this study suggested a convolutional neural network model for forecasting consumer churning behaviour. The fourteen thousand data examples obtained from one of the major cellular operators operating in Nigeria were divided into two datasets. Our model was created and put into use using the Python programming language and the Anaconda distribution. Our first preference was Jupyter notebook. A second CNN model, two multi-layer perceptron (MLP) models, and three additional models were created in order to conduct a comparable comparison. The accuracy rates for the MLP models are 81% and 89% for convolution neural network 1 and convolution neural network 2, and 80% and 81% for MLP 1 and MLP 2, respectively. When a company tries to bring in new clients, the costs are far higher than when it tries to lose old ones. All of their plans and strategies are structured to win over existing customers, lose new ones, and ultimately increase customer satisfaction. One of the biggest issues for this company, though, is user churn. It deprives a bank of a variety of income sources, including fees. Furthermore, in the Islamic banking system, the primary source of a bank’s income comes from consumer deposits. It might result in a bank withdrawing some of its deposits. If a bank ignores the loss of these two sources of income and the potential rise in reputational damage, it may be on the verge of failure. The current study offers a preliminary forecast of customer attrition for retail clients of an Iranian commercial bank. The accurate classification of clients in terms of churn rate is made possible by the use of cutting-edge data analysis algorithms to the transaction and operational data of bank customers. Thus, it can be designed to decrease the amount of user data by implementing appropriate solutions for each category.