Multi-Behavior Recommendation Model Based on Multi-Task Learning and Behavioral Dependence

As the scale of the e-commerce market continues to expand, the number of products is growing rapidly. How to efficiently obtain products that meet demand has become a major concern for e-commerce merchants and users. The product recommendation system can spontaneously find products that best match user preferences, and is an important method to solve the “information overload” problem of e-commerce platforms. As user needs are more closely integrated with recommendation models, recommendation performance is also facing higher requirements. How to further improve the accuracy of system recommendation is a continuing concern in academia and industry. Most of the existing recommendation algorithms based on deep learning only consider a single type of user behavior, ignoring the multiple types of behaviors that may occur when users browse products, such as viewing, adding to shopping cart and purchasing. These different behaviors contain a large amount of user preference information that is useful for recommendations. In order to consider the dependencies between various types of user behaviors in a fine-grained manner, we propose a multi-behavior product recommendation model, referred to as IBDM, that integrates behavioral dependencies into a multi-task learning framework. The IBDM model learns separate interaction functions for each behavior type, introduces a gating mechanism to adaptively learn the relationship between behaviors according to the actual situation, and introduces a tower structure for each behavior to output the user’s predicted value under that behavior. We conducted experiments on two real-world datasets to verify the recommendation effect of the IBDM model. Compared with the classic recommendation model, the IBDM model improved both HR and NDCG indicators.