MARKET SEGMENTATION OF STUDENT E-WALLET USERS USING K-MEANS CLUSTERING AND RANDOM FOREST

Abstract

The rapid growth of e-wallet use among students creates a need for computational workflows that explore variation in survey-based behavior and perception. This study aimed to segment student e-wallet users and identify features associated with the resulting clusters. Data were collected through an online questionnaire, and 569 valid responses were analyzed using 13 behavioral and perception features. K-Means produced two exploratory clusters: High-Perception Promo-Responsive Users (86.3%) and Lower-Perception Selective Users (13.7%). Because internal indices disagreed and the Silhouette score of 0.568 indicated only moderate separation, k = 2 was retained as an interpretability-oriented judgment; Silhouette and Davies-Bouldin supported k = 2, whereas Calinski-Harabasz and the elbow pattern supported k = 3. Random Forest was applied as a post-clustering interpretability tool to examine label reproducibility and identify influential features. Ease of use, trust, promotion and incentives, security, and risk perception were dominant, but the self-reported data require cautious interpretation.