Customer Churn Prediction in The Banking Sector on Imbalance Dataset

Customer churning is a significant concern of this digital age in every industry. As it is a highly cost sensitive phenomena for any business if any of their customer churned from their company to any other competitor. This is study is aimed to develop a highly accurate model by employing various machine learning algorithms for the prediction of the potential churners in the banking sector. To achieve this aim, a comprehensive amount of literature was reviewed for better understanding of the significant factors for such model designing. Data was prone to imbalance nature of class distribution because of that a couple of class balancing techniques which comprised upon SMOTE and ADASYN were employed. In the next phase, highly correlated features were extracted by using FCBF and RFE-FE and only those selected features were passed through different data mining techniques such as DT, RF and ANN. The results of the models‘ evaluation showed that model developed on the balanced dataset by ADASYN and having features from RFE technique perfumed best.