基于模糊逻辑的多代理推荐系统

Multi-agent recommendation system based on fuzzy logic

为模拟人的思维,向用户推荐高质量、个性化的电视节目,在TV-Anytime环境下提出一个基于模糊逻辑的多代理推荐系统.该系统包括模糊用户喜好档案、模糊筛选代理、模糊推荐代理、交互代理和档案修正代理.模糊用户喜好档案用于描述用户喜好和厌恶特征;模糊筛选代理基于模糊逻辑推理,将即将播放节目的元数据与用户喜好档案进行匹配,筛选出用户可能感兴趣的节目;模糊推荐代理根据预计的用户对推荐节目感兴趣程度排序,列出节目推荐表;交互代理收集用户清晰和隐含的反馈信息;档案修正代理根据用户的反馈,动态修正用户喜好档案中的特征参数值.实验结果显示该系统推荐效果可靠.

In order to imitate human＇ s thinking mode to recommend personalized TV program of high quality to users, a multi-agent recommendation system based on fuzzy logic is proposed under TV-Anytime environment. The system includes user＇ s fuzzy preference archive, fuzzy filtering/ranking agent, fuzzy recommendation agent, user interaction agent, and archive updating agent. The user＇ s fuzzy preference archive is used to describe both ＂like＂ and ＂dislike＂ features of users. Based on fuzzy logic inference, the filtering/ranking agent matches metadata of the incoming programs with user＇ s fuzzy preference archive, and selects programs that the user may be interested in. The fuzzy recommendation agent sorts the programs according to the degree of the expected user＇ s interest, and then presents recommended program list. The user interaction agent collects user＇ s both explicit and implicit feedback. The archive updating agent modifies the characteristic parameters in the user＇ s archive dynamically according to the user＇ s feedback. Results of the experiments show that the recommendation effect of the system is reliable.