Credit card customer churn is defined as the situation in which a bank’s customer stops using the bank's service and it leads to a potential loss of profit for the bank. Therefore, developing a customer churn prediction model for predicting customer likelihood of churning is essential for banks. This study aims to find out the key factor that influences credit card customer churn and to build a well performing model with high predicting accuracy. Finally, we can give some advice for banks to decrease customer churn. To achieve the goal, three models are employed, including neural network, logistic regression and decision tree model. The results indicate that all three models can be employed in different situations. However, the decision tree model has the best performance in both training and testing phases. The study further reveals that the top three factors influencing the likelihood of customer churn are total transaction count, total transaction amount, and changes in total transaction count. Based on the strengths and weaknesses of each model, this paper will also illustrate the different scenarios in which each model is most suitable.