基于视频类型和矩阵分解的加权推荐算法
首发时间:2019-05-17
摘要:在推荐算法领域,矩阵的奇异值分解(Singular Value Decomposition, SVD)算法是进行评分预测的有效工具。本文对基于SVD的评分预测算法进行了深入的研究,并结合视频类内容的特点,对该算法进行了改进,提出了一种基于电影类型和矩阵分解的推荐算法(Genre based SVD, GSVD)。GSVD算法利用电影类型这一现实场景中极易获取的属性信息,根据用户的历史评分数据学习用户对类型的喜好,利用类型生成一个电影的粗略评分预测,并将其作为SVD算法产生的精细评分的修正项,结合两者产生最终的预测评分。通过仿真结果表明GSVD算法表现出比传统的SVD算法更好的准确性。
For information in English, please click here
A Weighted Recommendation Algorithm based on Video Genre and Matrix Factorization
Abstract:In the field of recommendation algorithms, the Singular Value Decomposition (SVD) algorithm of matrices is an effective tool for scoring prediction. In this paper, the SVD-based scoring prediction algorithm is deeply studied. Combined with the characteristics of video content, a recommendation algorithm based on film genre and matrix decomposition (Genre-based SVD, GSVD) is proposed. The GSVD algorithm uses the information of movie genres that is easily acquired in the real scene, learns the user\'s preference for the movie genres according to the user\'s historical score data, and uses these information to generate a rough score prediction of a movie, and uses it as a fine score generated by the SVD algorithm. The corrections, combined with the two, yield the final predicted score. The simulation results show that the GSVD algorithm shows better accuracy than the traditional SVD algorithm.
Keywords: matrix factorization movie genre weightedalgorithm
基金:
引用
No.****
同行评议
共计0人参与
勘误表
基于视频类型和矩阵分解的加权推荐算法
评论
全部评论0/1000