Enhancing Product Recommendations and Sales Productivity with Advanced Deep Learning Models

Abstract

Recommender systems have revolutionized the way users discover products in online retail services. In the past, users had to invest significant time searching through long product lists, but recommender systems have made this process more efficient and valuable for users. This has led to increased revenue for service companies. In the retail industry, recommender systems have had two significant effects. Firstly, they have expanded the product portfolio available to customers by offering a greater variety of options beyond their previous purchases. Secondly, they have optimized the browsing and exploration time for customers and visitors, particularly benefiting less experienced visitors with targeted product recommendations.One of the most efficient and cutting-edge methods for addressing the challenges associated with diverse products sharing similar characteristics in the retail industry is the two-tower deep learning method. This method has demonstrated an impressive recall rate of 83% across all types of stores and products. In comparison, the fuzzy Technique for Order of Preference by Similarity to Ideal Solution method as an approach for making decisions that involve multiple criteria, achieved a recall rate of 70%, while the rule-based method achieved a recall rate of 52%. These results highlight the substantial superiority of the proposed method in artificial intelligence for retail engineering applications.

Keywords: recommender system, Deep learning, Retail service, Rule-based method