Sales Predictive Model with Customer Segmentation Integration

Predictive analytics is a fundamental tool for optimizing business decisions in contexts such as e-commerce and marketing. This paper proposes a sales prediction model that integrates customer segmentation information generated through clustering techniques. Three customer segments are identified from an e-commerce transaction dataset: the first characterized by customers who make recent, low-volume purchases; the second group related to customers with high historical investment but no recent activity; and the third group, customers who purchase frequently at medium levels. This segmentation was an additional predictor variable in the regression model to predict future sales. Two clustering algorithms, DBSCAN and K-means, were tested and evaluated with the Silhouette metric, while the prediction model using Random Forest was evaluated with conventional metrics such as Root Mean Square Error and R-Squared. The results indicate that the integration of clusters into supervised models can enhance prediction models, demonstrating good performance.