Abstract: In order to recommend items to a group of users as well as improve their information searching efficiency, a novel group recommendation method based on optimized collaborative filtering and the median-based weighted average is proposed. The method takes items’ rating similarity and type similarity into consideration, integrates item similarity and user similarity to predict the values of items which users have not yet rated. Then it uses median-based weighted average strategy to aggregate the group of users’ ratings, taking the?effects?when?users?rating?into?consideration. In the end two experiments to?predict?items’?ratings and?integrate?users’?ratings?are?given?out?respectively. The?results show?that two algorithms are better than traditional ones in terms of accuracy, indicating that the strategy proposed is valid.

Key words: group recommendation, item similarity, user similarity, median, weighted average

摘要： 为面向群体用户提供推荐，提高群体用户的信息搜索效率，提出了一种新颖的基于优化协同过滤与中位数加权平均的群推荐方法，综合考虑了项目的评分相似性与类型相似性，通过集成项目相似性与用户相似性预测出群体用户对项目的评分；在集结群体用户评分时，采用基于中位数的加权平均集结策略消除个别用户评分差异过大带来的影响，综合考虑群体用户在评分过程中的作用。通过预测项目评分实验与集结用户评分实验，结果表明，用新方法得到的准确率均高于常用的传统方法，从而表明该方法是有效的。

关键词: 群推荐, 项目相似性, 用户相似性, 中位数, 加权平均策略