关键词: Web日志, 用户群偏爱度, 模糊聚类算法, 混合阶Markov模型, 预测

Abstract: In the Web environment，measuring users’ browsing patterns can benefit the improvement of framework of Web sites.Mining and measuring Web logs are able to identify users’ navigation patterns models，Web-masters can apply these models for studying users’ access favoritism to improve site organization and analyze the effects of changes to their Web sites.So，in this paper，a new conception of user group favoritism is proposed，and Fuzzy Clustering algorithm based on User Group（UGFC） is used，and then，according to clustering result—user group having similar access habit，user group favoritism is measured，based on which the Hybrid-Order Markov Model（HOMM） is used to predict.This new prediction metrics approach（UGFC-HOMM） shows that it is superior to Traditional Markov Model（TMM） in users’ access prediction.Three evaluation metrics are applied to evaluate performance of prediction，namely，precision，coverage and runtime.

Key words: Web logs, user group favoritism, fuzzy clustering algorithm, hybrid-order Markov model, prediction