|
Behavior and characterization of users is an important issue in the design and maintenance of websites. Analysis of the Web access logs can offer deep comprehension to the Web usage and facilitate the Web personalization. This paper presents a method for Web user categorization from Web log files. A link graph is constructed firstly after the data cleaning and transaction identification with original Web log data. The similarity between two pages in the link graph are defined which is based on the page link information and by merging the similar pages into a page class the link graph is compressed. Finally we use the FDOD, a measure of discrepancy between ordered sequences, to categorize the link paths into several classes. The experiments on an actual Web site log data are tested and the results indicate that the approach proposed have advantages of easy-to-use, fast-response and good accuracy. |
|
Keywords:Web usage mining, categorization, link path, FDOD |
|