Mining Website Log to Improve Its Findability

Under the network environments with large amounts of digitalized data, websites are the information strongholds that institutions, organizations or enterprises must set up for their specific purposes. No matter how they have been built, websites should offer the capability that users can find their required information quickly and intuitively. Surfing around the library websites, the website logs always keep tracks of users' factual behaviors of finding their required information. Thus we can apply data mining techniques possibly to explore users' information seeking behavior. Based on these evidences, we attempt to reconstruct the websites to promote their internal findability. In this paper, we proposed a heuristic algorithm to clean the website log data, to extract user sub-sessions according to their respective the critical time of session navigation, and to calculate each sub-session's the threshold time of target page with different weights to determine its navigating parent page. We utilized the alternate parent pages of weights to reconstruct various websites. We conduct task-oriented experiments of 4 tasks and 25 participants to measure the effects of their findability respectively. By the analysis of variance on time to complete the tasks, the result has shown that the reconstructed website has better findability performance.