摘要
Abstract
Recommendation systems based on user behavior modeling are widely used in recall,sorting,and other stages,including sequential recommendations and session recommendations.In the sequence recommendation,some behavior records that are not related to the next behavior may be introduced due to excessively long sequence settings.However,the session-based recommendation focuses on medium and short-term recommendations,so it has limitations in capturing long-term or general interests.In view of the above,a sequence recommendation system(T-BOI)that integrates temporal and behavioral order information has been proposed,making it suitable for long-term and short-term interest recommendations.In the proposed method,the feature representation unit module,behavior weight unit module,behavior sequence representation model,and behavior category output unit module in T-BOI are used to process and obtain the final prediction result.The proposed method is compared with some advanced models.The validation on public datasets shows that the recommendation system has good recommendation effect.关键词
序列推荐/推荐系统/多行为推荐/长短期偏好/位置编码/时间信息Key words
sequence recommendation/recommendation system/multi behavior recommendation/long and short-term preference/position encoding/time information分类
电子信息工程
