Desarrollo de un sistema de recomendación de compra utilizando algoritmos de aprendizaje automático de sugerencia para la promoción efectiva de combos y ofertas

Abstract

CONDITION FOR PUBLICATION OF PROJECT. In the retail sector, improving product offerings is essential for increasing sales and enhancing the customer experience. This study develops a product recommendation system aimed at commercial and sales teams, with the goal of suggesting product combinations that improve profitability. Three machine learning algorithms were implemented: K-Nearest Neighbors (KNN), Alternating Least Squares (ALS), and Neural Collaborative Filtering (NCF), evaluating their performance using Precision, Recall, and F1-Score. Historical purchase data was analyzed to identify consumption patterns and segment products. The models were then trained and evaluated in different scenarios to determine the best-performing model for generating recommendations. The results show that NCF achieved the best performance, providing more accurate suggestions. Additionally, an interactive application prototype was developed, allowing commercial teams to select products and receive real-time recommendations. This approach facilitates promotion planning and improves data-driven decision-making. It is concluded that the developed recommendation system serves as a valuable tool for product management in retail, enhancing the customer experience and the efficiency of commercial strategies. Keywords: Recommendation systems, Retail, Machine learning, Data science.