An Explainable Predictive Analytics Framework for Customer Behavior Forecasting, Personalized Marketing, and Intelligent Strategic Decision-Making

Customer behavior analytics has become increasingly important for supporting personalized marketing and strategic business decisions. Explainable predictive approaches can help identify customer patterns while maintaining transparency in model interpretation. This study aimed to develop an explainable predictive analytics framework for customer behavior forecasting, personalized marketing, and strategic decision-making. A secondary customer segmentation dataset containing 53,503 observations and 20 variables was analyzed. Descriptive statistics, chi-square tests, Cramér’s V, Pearson and Spearman correlations, and multiclass classification models were applied. Multinomial Logistic Regression, Decision Tree, and Random Forest models were compared using accuracy, precision, recall, and F1-score. Feature importance and permutation importance were used for explainability. Customer segmentation was significantly associated with several demographic, behavioral, service-related, and preference variables, although effect sizes were weak. Multinomial Logistic Regression achieved the highest overall accuracy, while Random Forest showed the strongest macro F1-score. Coverage amount, premium amount, income level, age, and purchase-related temporal variables were among the most influential predictors. Segment-specific differences were also observed in communication preferences, contact timing, and customer preferences. Explainable predictive analytics provided useful insights for customer segmentation and personalized marketing, but limited predictive separation indicated that richer longitudinal and transaction-based data would be needed to improve forecasting performance and strategic applicability.