Customer Segmentation of Mobile Banking Users Using Feature Engineering and K-Means Clustering

The increasing adoption of mobile banking has necessitated deeper insights into user behavior to enable banks to design personalized and targeted marketing strategies. This study aims to segment mobile banking customers based on their transaction patterns, specifically in the purchase of prepaid mobile credit and internet packages, using feature engineering techniques and the K-Means clustering algorithm. A dataset comprising over one million transactions from a regional bank in North Sumatra, Indonesia, was analyzed. Behavioral and time-based features were extracted to capture customer activity levels, transaction values, temporal preferences, and product usage. The Elbow Method identified five optimal clusters, each representing unique user profiles, including occasional users, regular low-value users, premium users, heavy users, and moderate-consistent users. Findings indicate strong operator loyalty and consistent transaction timing across segments, especially in early-month activity. The results offer practical implications for financial institutions seeking to enhance customer engagement, retention, and service personalization through behavior-based segmentation strategies. This study also contributes methodologically by showcasing the utility of unsupervised machine learning in deriving customer insights from transactional data without relying on sensitive demographic information.