Diversified Recommendation Algorithm Based on Penalty Factors

Abstract

Abstract Personalized recommendation system is a technology that uses user behavior and preference information to provide personalized recommendations for users. With the development of the Internet and the era of information explosion, personalized recommendation systems have been widely used in e-commerce, social media, music, video and other fields. Negative feedback-based user diversity recommendation algorithms aim to provide richer and diverse recommendation results to satisfy users' different interests and needs. Traditional recommender systems usually face the problems of over-personalization and user information bubbling because they mainly rely on positive feedback signals (user clicks, purchases, etc.), which may lead to users being recommended similar content and ignoring potential diversity. At the same time, due to the influence of social attributes, people tend to be influenced by mainstream elements, which leads to the problem of "information cocoon" when recommending, although it can still recommend content for the user's satisfaction, but does not really take into account the user's own interests and needs. This paper combines a series of commonly used recommendation algorithms, such as itemCF, userCF, CB, etc., to take into account the social type at the same time, more consideration of the user's own personalized recommendations, and the introduction of "negative feedback" mechanism to reduce the popular labels on the impression of the user's real interests, to further explore more innovative and interesting content, to achieve the effect of The effect of thousands of people is realized.