Abstract: With the rapid development of the Internet and the wide application of web-based system，more and more research works are come out with the web mining.Data mining in web log which focuses on the discovery and analysis of useful information in web log has been a hot spot.Many data mining methods based on association rule have been proposed.In this paper，the rough set theory based on discernibility matrix is applied to derive association rule from web log，and the derived association rule set is used to predict user behavior.From experiment results，it is shown that this algorithm is valid and practicable.

Key words: rough set, user’s behavior prediction, web log mining, association rule, discernibility matrix

摘要： 随着互联网的飞速发展和Web应用系统的广泛应用，Web挖掘得到了人们越来越多的研究。从Web日志中发现和分析出用户的有用信息的Web日志挖掘已成为研究热点。很多基于关联规则的方法已经被应用于Web挖掘中。运用基于差别矩阵的粗糙集提取Web日志中的关联规则，并将生成的关联规则集用于用户行为的预测。实验结果说明该方法的有效性和实用性。

关键词: 粗糙集, 用户行为预测, Web日志挖掘, 关联规则, 差别矩阵