Analysing consumer behaviour using machine learning to enhance customer cross-selling recommendation

Abstract

The growing complexity of consumer behavior in the retail industry necessitates advanced solutions for optimizing cross-selling and inventory management. Traditional methods often lack scalability, personalization, and adaptability. This study develops a hybrid machine learning recommendation system using customer demographics, purchase history, and product data to enhance cross-selling strategies. By integrating K-means clustering, DBSCAN, Collaborative Filtering, and Content-Based Filtering, the system delivers personalized product suggestions and identifies slow-moving inventory. Deployed through a Streamlit interface, the hybrid model achieved improved precision (20.7%), recall (67.2%), and NDCG (0.508). Statistical testing confirmed these gains were significant, and segmentation was shown to improve recommendation quality. The system offers practical benefits in sales growth and stock optimization, with future potential for real-world deployment to drive sustainable, personalized retail experiences. Keywords: Cross-selling, Machine Learning, Customer Segmentation, Inventory Optimization, Retail Analytics