Web Mining and Social Networking: Techniques and ... - tud.ttu.ee

136 6 Web Usage MiningFig. 6.7. An example of index page derived by PageGather algorithm[204]to T.(b) Else, either go back (backtrack) to the parent of C with some possibility, or cease withsome possibility.For a set of target pages (T 1 ,T 2 ,···,T n ), the search pattern follows the similar procedure,but after the user identifying T i , it continues searching T i+1 . In this scenario, the hardest taskis to differentiate the target page from other pages by simply looking at the Web log. In thisstudy, the authors claimed that the target pages could be separated out based on either whetherthe targets are content pages or index (or navigational) pages, or setting a time threshold. Forthe former case, the content pages are most likely to be the target pages for a user. Givena website of portal site, where there is not a clear separation between the content and indexpages, resulting in the difficulty in judging the content pages, counting the time spent on aspecific page will provide an useful hint to judge this. Here it is known that the user spentmore time than the time threshold are considered the target page.To identify the backtrack points, the Web log is analyzed. However, the browser cachingtechnology brings in unexpected difficulty in differentiating the backtrack points, otherwise,the phenomenon of a page where previous and next pages are the same gives the justificationto it. In this work, rather than disabling the browser caching, a new algorithm of detectingbacktrack points was devised. The algorithm is motivated by the fact that if there is no linkbetween P 1 and P 2 , the user must click the “back” button in the browser to return from P 1 to P 2 .Therefore the detection of the backtrack points is becoming the process of detecting whetherthere is a link between two successive pages in the Web log. To do this, the authors built a