基于时间效应的多准则推荐算法

Multi-criteria Recommendation Algorithm Based on Time Effect

近年来,推荐系统被广泛认为是解决＂信息过载＂及＂信息迷航＂的一个有效工具。多准则评分比单一整体评分具有更为丰富的用户个性化偏好信息,但传统的多准则推荐系统研究未考虑到用户兴趣漂移的情况。针对这一问题,本文将时间信息与基于用户的多准则协同过滤算法相结合,在多准则算法中引入基于遗忘规律的艾宾浩斯遗忘曲线拟合用户兴趣漂移,修正用户之间的相似度计算结果。实验结果表明,与传统的多准则协同过滤算法相比,本文提出的算法具有较高的准确度。

In recent years,the recommendation system has been widely considered as an effective tool for solving ＂information over-load＂ and ＂information trekking.＂ The multi-criteria rating has more user personalized preference information than a single overallrating,but the traditional multi-criteria recommendation system does not consider the situation of user interest drift.In order tosolve this problem,this paper combines time information with user-based multi-criteria collaborative filtering algorithm.In themulti-criteria algorithm,Ebbhauser forgetting curve based on forgetting rule is introduced to fit user interest drift,and similaritybetween users is corrected.Experimental results show that compared with the traditional multi-criteria collaborative filtering algo-rithm,the proposed algorithm has a higher accuracy.