Data mining methods and their application in solving credit card customer churn situations

Bc. YuFeng Gao

Master's thesis

Abstract:

The theoretical part of the paper first I start with the definition of big data, I introduce the basic concepts and principles of data mining algorithmic models in big data, then I focus on the development of statistical data processing and the basic methods of statistical and predictive analysis. After that I introduce three different data mining models used to process big data: logistic regression model, decision tree CHAID model and decision tree C&D model. they are after that I will use to predict and determine credit card customer churn. At the end of the theoretical part, I briefly introduce the development and use of credit cards and the general definition of customer churn behavior, and I provide a preliminary analysis of the comprehensive factors influencing customer churn behavior. In the practical part of the work, I deal with a large dataset of 10,127 objects, a collection containing information related to credit card customers, which is placed on the Kaggle online platform for educational purposes. The practical work involved creating appropriate classification algorithms and using analytical tools to identify the factors that influence credit card customer churn. Based on these factors, logistic regression models and two different decision tree models were constructed to predict customer churning behavior. Finally, data analysis tools and testing of test datasets were used to validate the predictive power of the models. Three data prediction models were built using logistic regression and decision tree algorithms encapsulated in IBM SPSS to build data models based on a dataset of credit card customer churn. Their prediction quality was checked by using confusion metrics and the area under the receiver operating characteristic (ROC) curve (AUC). By comparing the three prediction models, a data model with 91% prediction accuracy is established, which can provide a reliable reference for banks' credit card customer attrition prediction.

Keywords

Statistical analysis, Predictive models, Binary logistic regression, Decision tree models, Big data, Customer churn

Language used: English

Date on which the thesis was submitted / produced: 4. 4. 2024

Thesis defence

• Date of defence: 27. 5. 2024
• Supervisor: Ing. Tomáš Hlavsa, Ph.D.
• Reader: Michal Kadlec, externi

Czech University of Life Sciences Prague

Faculty of Economics and Management

Master programme:
Economics and Management