Earticle

초록

영어

Commercial banks are increasingly expanding intermediary services, such as mutual funds and bancassurance, to reduce reliance on interest-margin-based profit models. However, identifying customers with a high probability of purchasing such products remains a significant challenge for retail banks. Although machine learning (ML) has been widely applied in banking, most existing studies focus on risk management tasks, while relatively limited attention has been paid to customer targeting, product adoption, and cross-selling for intermediary financial services. In addition, highly accurate ML models often lack the interpretability required for marketing management. To address these gaps, this study proposes an interpretable ML framework for predicting customer purchasing behavior for bank intermediary products. Using survey data from 368 retail banking customers in Hangzhou, China, this study constructs a multidimensional customer profile including demographic, financial, investment-related, behavioral, health-related, and psychological variables. Several classification models are compared, including Logistic Regression (LR), Naïve Bayes (NB), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Neural Network (NN). The results show that XGBoost achieves the highest performance among the compared models. Furthermore, SHapley Additive exPlanations (SHAP) analysis and Partial Dependence Plots (PDPs) are employed to interpret the prediction results. The interpretability results reveal that investment experience and deposit balance are the most important predictors of purchasing behavior. Behavioral engagement variables, such as transaction frequency, credit card use, and mobile banking activity, also contribute to customer classification. The interaction analysis further indicates a compensatory pattern: active daily banking engagement can increase purchase probability even among customers with no clear investment preferences. This finding suggests that purchase decisions are shaped not only by financial capacity and prior investment experience, but also by customers’ ongoing interactions with banking services. Theoretically, this study extends ML applications in banking from risk management to proactive cross-selling. Practically, the findings provide implications for customer segmentation, precision marketing, and intermediary product recommendation. Overall, this study shows that interpretable ML can deliver both predictive accuracy and managerial insight in retail banking.

목차

Abstract
1. Introduction
2. Literature Review
2.1. Traditional Customer Classification in Banking
2.2. Machine Learning Applications in Banking
2.3. Research Gaps
3. Research Overview
3.1. Data Source and Processing Methodology
3.2. Interpretability Analysis
4. Experimental Results and Analysis
4.1. Logistic Regression Baseline Analysis
4.2. Performance of the Models
4.3. Interpretability Analysis with SHAP and PDPs
5. Discussion
6. Conclusion
6.1. Main Findings
6.2. Theoretical Implications
6.3. Practical Implications
6.4. Future Research Directions
References
Appendix

키워드

저자

• Wang, Xianzhe [ Department of Business Administration, Chungnam National University, Republic of Korea ]
• Lee, Jun-Ho [ Department of Business Administration, Chungnam National University, Republic of Korea ] Corresponding Author

참고문헌

발행기관

간행물

표지보기 관심저널 등록

• 간행물명

  Journal of China Studies

• 간기

  계간

• pISSN

  1975-5902

• eISSN

  3022-5590

• 수록기간

  2006~2026

• 등재여부

  KCI 등재

• 십진분류

  KDC 912 DDC 951