Modeling financial literacy through explainable machine learning and behavioral segmentation in emerging economies

Digital financial services are changing how people behave financially but understanding how individuals use financial technology remains limited in emerging economies. This study investigates financial literacy and technology-driven financial behavior using machine learning analysis of 1,067 adults in Bangladesh. Traditional demographic methods fail to capture the complexity of digital financial behavior, which hampers the design of effective interventions. By applying Random Forest, XGBoost, and k-means clustering validated with a silhouette score of 0.42, Davies-Bouldin Index of 1.08, and Calinski-Harabasz Index of 287.3, we achieved moderate predictive performance with an F1-score of 0.52, a 58% improvement over random guessing, and gained important insights into technology-influenced financial behavior. SHAP analysis identified institutional trust, digital comfort, and income as key predictors, with trust showing an importance value of 0.18 compared to education at 0.09, challenging typical demographic assumptions. Three distinct behavioral groups emerged: Digitally Literate Planners (34% of the sample, average financial knowledge 7.8/10), Informally Active but Underskilled (41%, knowledge 5.4/10), and Digitally Excluded Traditionalists (25%, knowledge 3.1/10). Results show that rural users had 7.3% higher financial literacy than urban users, while formal education did not significantly impact literacy levels. Weak correlations between financial knowledge and actual behaviors, with coefficients below 0.10, suggest that understanding alone does not lead to effective financial practices. The findings provide frameworks for tailored interventions based on behavioral types instead of stereotypes, including resource allocation of 60% for support, 30% for transitional programs, and 10% for advanced services, to improve intervention effectiveness for diverse users in developing economies. • Rural participants outperformed urban residents in financial literacy by 7.3%. • Institutional trust predicts literacy better than formal education level. • Three behavioral clusters identified requiring differentiated interventions. • Knowledge-behavior gap reveals understanding alone does not ensure action. • ML with SHAP analysis provides interpretable insights for policy design.