Experimental Verification of the Dependence Between the Expected and Observed Visit Rate of Web Pages

摘要

This paper is focused on a utilization of the web usage mining and web structure mining methods. We tried to answer the question if the expected visit rate of individual web pages correlates with the observed visit rate of the same web pages. We used web server log files as a data source. We applied several log file pre-processing methods to identify the user sessions on different levels of granularity. We found out that the quality of acquired knowledge about the users' behaviour depends on the method of the session identification. We have experimentally proved a higher dependence between the observed and expected visit rates of the examined web pages in well-prepared files with identified user sessions. We found out statistically significant differences between PageRank and a real visit rate in the files with application of more advanced methods of session identification.