Evaluating the performance of machine learning algorithms in predicting the best bank customers

Abstract

The best customer refers to the potential interaction of customers with the company during certain time periods. When companies understand the best customer and realize that the best customer can provide customized services for different customers, then they will achieve effective customer relationship management. This research is focused on the banking industry and systematically integrates data mining techniques and management topics to analyze the best customers. This study first uses the fuzzy hierarchical analysis method to weight the existing variables and then examines the DFMT model as an input to the k-means technique for clustering customers based on the desired criteria in the DFMT model. By using the proposed scoring model, it starts forming a customer value pyramid and categorizes customers into 4 value spectrums. Finally, in order to analyze the classes obtained from the customer value pyramid and implement the learning process from the available data, it uses the tenor classification techniques of decision tree, support vector machines and random forest along with the six characteristics and among They introduce the most appropriate model-characteristic based on available criteria.