Improving bank customer churn prediction with feature reduction using GA

Customer churn is one complex problem that all banking institutes struggle to deal with. The increased competition in the industry is forcing the banks to maintain customer loyalty by providing apt services to the demanding customers. Analysing the customer churn data will enable the banks to plan their services effectively and depend on various machine learning techniques to predict this churn. However, the high data dimension in these datasets adds adverse effects to this algorithm's performance. This work proposes a way to improve the prediction accuracy of high-dimensional data with the help of feature selection using Genetic Algorithm (GA). This model identifies the machine learning algorithms which are sensitive towards the feature selection and which are not. The model is tested on a primary dataset and the results are analysed. This research also analyses those features which are the most important in predicting customer churn while applying the GA-optimized classification models. This assists the banking institutions to revise their marketing strategy so that they can attract more customers and also reduce customer attrition by providing the apt services to them.