摘要
为解决协同过滤推荐中"稀疏"和"冷开始"问题,提高推荐精度,提出了基于隐式评分的推荐系统。首先建立项档案,采用BP神经网络模型分析用户的导航模式和行为模式,对已点击项进行预测评分,建立用户主观评价模型和用户偏好档案;然后预测用户对未点击项评分,形成比较稠密的用户预测评分矩阵,采用协同过滤推荐技术,产生有效推荐;最后提出基于项特征的谈判模型和谈判策略,支持对推荐结果的解释和客商之间的讨价还价。
Recommendation system based on implicit rating was proposed to improve the precision and solve the problems of ＂scarcity＂ and ＂cold-start＂. Firstly, this research set up the items＇ profiles, and adopted the BP neural network to analyze the guiding model and behavior model of the users, gave the forecast rating for the hit items and set up subjective evaluation model and the profiles of preference for the users. Then it forecasted the rating of the non-hit items and formedthe intense rating matrix of user foreast item. After that, it produced the effective recommendation through the adoption of collaborative filtering recommendation algorithm. Finally, the model of negotiation and strategy based on item＇s characteristics were brought out for recommendation result, which can explain the result and support the bargaining of both sides.
出处
《计算机应用》
CSCD
北大核心
2009年第6期1585-1589,共5页
journal of Computer Applications
基金
湖北省教育厅项目(D200819022008d095)
关键词
协同过滤
隐式评分
推荐系统
策略
collaborative filtering
implicit rating
recommendation system
strategy
