Research On Predicting Commercial Bank's Credit Card Customer Loyalty

With the rapid development of domestic retail banking industry, the credit card's issuing amount is increasing at a high speed annually. The credit card customers are definitely to be the main part of banks'profit. Maintaining and enhancing credit card customer loyalty is the major objective of bank's customer relationship management system, it has also been a hot issue among researchers. Researches on customer loyalty home and abroad belong to empirical research in recent years, mainly discussed the key determinants and influential factors of customer loyalty under various industries. Predicting research on customer loyalty is still a comparatively new field, especially based on the transactional data.This paper makes some creative work in the field of customer loyalty prediction based on the real credit card transactional data of a domestic commercial bank, and proposes a customer loyalty predicting model based on neural network technique.The main work and contributions of this paper are as follows:1. After making a fully investigation, this paper accomplishes the certain domestic commercial bank's real credit card transactional data preprocessing work including data cleaning and transactional record combining; deeply analyzes the deficiency of the questionnaire application, proposes a customer loyalty evaluation system based on customer's objective features.2. This paper proposes a customer loyalty predicting model based on back propagation (BP) neural networks, modifies the model parameters, and compares different training algorithms; validates this predicting model on the credit card transactional data, and the result shows this predicting model gets a good predicting accuracy.3. This paper applies another three different data mining techniques to build up customer loyalty predicting modeless, including Probabilistic Neural Networks, Support Vector Machine and Decision Tree; deeply compares the four predicting model, and illustrates that the customer loyalty predicting model based on BP neural network is the best from both point of entire testing data predicting accuracy and high loyalty customers'predicting accuracy.4. This paper accomplishes the rule extraction work on the customer loyalty predicting model based on BP neural network, applies BIO-RE algorithm to extract a series of customer loyalty predicting rules, enhances the explanation of the predicting model, and these rules are somewhat valuable to bank managers.