Loading...
Paper Type
ERF
Abstract
Churn prediction is the process of using transaction data to identify customers who are likely to cease their relationship with a company. To date, most work in churn prediction focuses on sampling strategies and supervised modeling over a short period of time. Few have explored the area of mining customer behavior pattern in longitudinal data. This research developed a dynamic approach to optimizing model specifications by using time-series predictors, multiple time periods, and rare event detection to enable accurate churn prediction. The study used a unique three-year dataset consisting of 32,000 transaction records of a retail bank in Florida, USA. It uses trend modeling to capture the change of customer behavior over time. Results show that data from multiple time periods helped to improve model precision and recall. This dynamic churn prediction approach can be generalized to other fields for which mining long term customer data is necessary.
Recommended Citation
Leung, Hoiyin Christina and Chung, Wingyan, "A Dynamic Classification Approach to Churn Prediction in Banking Industry" (2020). AMCIS 2020 Proceedings. 28.
https://aisel.aisnet.org/amcis2020/data_science_analytics_for_decision_support/data_science_analytics_for_decision_support/28
A Dynamic Classification Approach to Churn Prediction in Banking Industry
Churn prediction is the process of using transaction data to identify customers who are likely to cease their relationship with a company. To date, most work in churn prediction focuses on sampling strategies and supervised modeling over a short period of time. Few have explored the area of mining customer behavior pattern in longitudinal data. This research developed a dynamic approach to optimizing model specifications by using time-series predictors, multiple time periods, and rare event detection to enable accurate churn prediction. The study used a unique three-year dataset consisting of 32,000 transaction records of a retail bank in Florida, USA. It uses trend modeling to capture the change of customer behavior over time. Results show that data from multiple time periods helped to improve model precision and recall. This dynamic churn prediction approach can be generalized to other fields for which mining long term customer data is necessary.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.