| Summary: | 碩士 === 國立臺灣大學 === 資訊管理研究所 === 89 === The managers of web sites want to meet the objectives of customer-centric and one-to-one marketing, they can discover the important data patterns from the log files that stored at the web server.
In this thesis, we propose four types of algorithms to analyze the user access behavior:
Type 1: Explore the frequent path traversal patterns in a certain time period. That is, explore the paths that the users access frequently in a certain time period.
Type 2: Explore the cyclic path traversal patterns under different time units. That is, explore the cyclic paths that the users access frequently under different time units, and find the time change of path traversal patterns.
Type 3: Explore the frequent path traversal patterns by a moving/sliding window. That is, explore the paths that the users access frequently by a moving/sliding window, and provide a better data mining method for analyzers.
Type 4: Explore the frequent path traversal patterns for an individual user. That is, explore the paths that the individual user accesses frequently by analyzing the patterns of his/her navigation paths.
In order to show the efficiency and effectiveness of our algorithms, we perform two series of experiments. One is made on the synthesized data. The other is made on the log files of two-year time period from DGBASEY (Directorate-General of Budget, Accounting and Statistics, Executive Yuan, Republic of China, Taiwan). We do discover the useful access patterns for improving the structure of web pages via our experimental results.
|
|---|
