Prediction of Churning behavior in Telecom using Machine Learning

For any industry to flourish, the key role is played by their customers. Growth of industry is directly proportional to customer-database. It is always said and believed throughout that the process of retention of customers is easier than finding new customers especially in the case of the Telecom industry. Thus, the process of detecting or predicting those customers who might leave the organization in the near future is called churn behavior prediction of customers and this process is most important in today’s technological world. To get this most wanted information, we have incorporated a model that detects customer churn behavior. This model is constructed by the help of two supervised machine learning models, Random Forest and Artificial Neural Network. The pre-stored data is preprocessed using one hot encoding technique, normalization techniques and analyzed using k-means clustering. This data is applied on Random Forest in order to obtain important features. Top twelve important features are retained and thus refined dataset is fed into ANN model to obtain 83.8% accuracy. These important features have a feature importance score that tells where the organization is going wrong in retaining its customers.