基于关联规则和Multi-Agent的个性化信息推荐系统研究

PIRS Based on the Association Rule Mining and Multi-Agent

为提高网络信息激增中个性化信息推荐的有效性和智能性,将关联规则技术和Multi-Agent技术应用到个性化信息推荐中,设计一个通过对用户日志挖掘以产生个性化信息推荐的系统PIRS。该系统包含6个不同层次具有独立功能而又相互关联的Agent任务模块,引入多个Agent收集和分析用户信息来学习用户的兴趣和行为,体现个性化信息推荐的智能性;利用PIRAgent在用户日志中进行挖掘时,采用的关联规则挖掘方法是基于位对象技术和改进的FP-Tree构造方法,提高系统推荐效率。

In order to improve effectiveness and intelligence of personalized information recommendation with the quickly increasing Web information, a system named PIRS （Personalized Information Recommendation System）is designed, and association rule mining technology and Multi-Agent technology is applied to PIRS, which can offer personalized information recommendation services for users by mining user action logs. PIRS consists of six different task modules, each of which has independent function but related to each other. Many agents are introduced to collect and analyze user information so as to learn the users＇ interests and actions, so the intelligence of the system based on multi-agent is proved. When user action logs are mined by PIRAgent, association rule mining algorithm based on the bit objects technology and the improved constructing method of FP-Tree （Frequent Pattern Tree） is used, and the system recom- mendation efficiency is improved.