基于评论挖掘的改进的协同过滤推荐算法  

Improved Collaborative Filtering Recommendation Algorithm Based on Comments Mining

作　　者：王全民[1] 王莉[1] 曹建奇[1] 

机构地区：[1]北京工业大学计算机学院,北京100124

出　　处：《计算机技术与发展》2015年第10期24-28,共5页Computer Technology and Development

基　　金：国家自然科学基金资助项目(61272500)

摘　　要：随着因特网的飞速发展,电子商务网站为人们提供了越来越多的选择,随之而来的信息过载和信息迷失问题日益严重,个性化推荐系统的出现极大地改善了这一情况。协同过滤是目前主流的推荐算法,但随着用户物品数目的日益增多和系统规模的不断扩大,用户-物品评分矩阵存在着严重的稀疏性等问题,导致推荐系统的推荐质量严重下降。针对此问题,文中提出了一种改进的协同过滤推荐算法,将评论挖掘技术引入协同过滤算法中,量化物品在各个特征上的分数,然后结合物品特征和用户评分共同计算物品相似度,将得到的物品预测评分填充用户-物品评分矩阵,最后结合基于用户的协同过滤思想对用户产生推荐。实验结果表明,改进的协同过滤推荐算法提高了推荐结果的精确度。With the rapid development of the Interact, electronic commerce website provide more choice for people, but information over- load and information lost problems become increasingly serious, the personalized recommendation system has greatly improved this situa- tion. Collaborative filtering recommendation algorithm is a popular recommendation algorithm, but with the increasing number of items and users and the continuous expansion of the system, the serious user-item rating matrix sparse problem leads to a lower recommended quality. In order to solve this problem, put forward an improved collaborative filtering recommendation algorithm, which introduces com- ments mining technology into collaborative filtering algorithm to get the item score on each feature, and then combine the feature of items and the user score to calculate the item similarity, fill the predicted rating score into the user-item rating matrix, finally recommend to the user based on the user-based collaborative filtering ideas. The experimental result shows that the improved collaborative filtering recom- mendation algorithm improves the precision of the recommendation results.

关 键 词：评论 协同过滤 相似度 推荐算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]