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

134 6 Web Usage MiningWith the propagation and popularity of search engine, Web Usage Mining has attracted a lotof interests on how to facilitate the search performance via learning usage knowledge. Moreover,the research of Web community was greatly benefited from the advance of Web UsageMining. In this section, we will review some studies carried out in this area.6.5.1 Mining Web Logs to Improve Website OrganizationA good design and organization of a website is essential in improving the website’s attractivenessand popularity in Web applications. A well designed and organized site is often a basicrequirement for securing the success of a site. However, it is not a easy task for every Webdesigner to satisfy the aims initially designated. There are many reasons, which are associatedwith the above disappointments. First, different users have their own navigational tasks sofollowing different access traces. Second, even the same user may have different informationneeds at different times. Third, the website is not logically organized and the individual Webpages are aggregated and placed in inappropriate positions, resulting in the users uneasily locatingthe needed information; and furthermore, a site may be designed for a particular kindof use, but be used in many different ways in practice; the designer’s original intent is not fullyrealized. All above mentioned reasons will affect the satisfactory degree of a website use. Bydeeply looking into the causes of such reasons, we can intuitively see that it is mainly becausethe early website design and organization only reflects the intents of website designers or developers,instead, the user opinions or tastes are not sufficiently taken into account. Inspiredby this observation, using Web Usage Mining techniques is intuitively proposed to addressthe improvement of website design and organization. Essentially the knowledge learned fromWeb Usage Mining is able to reveal the user navigational behavior and to benefit the siteorganization improvement by leveraging the knowledge.Adaptive Web SitesIn [204, 203] the authors proposed an approach to address the above challenges by creatingadaptive websites. The approach is to allow Web sites automatically improve their organizationand presentation by learning from visitor access patterns. Different from other methods such ascustomized Web sites, which are to personalize the Web page presentation to individual users,the proposed approach is focused on the site optimization through the automatic synthesisof index pages. The basic idea of the proposed is originated from learning the user accesspatterns and implemented by synthesizing a number of new index pages to represent the useraccess interests. The approach was called PageGather algorithm. Since the algorithm needsonly to explicitly learn usage information rather than to affectively make destructive changesto the original Web sites, the author claimed that the strength of the approach is nondestructivetransformation.The PageGather algorithm is based on a basic assumption of visit-coherence: the pagesa user visits during one interaction with the site tend to be conceptually related. It uses clusteringmining to find the aggregations of related Web pages at a site from the access log. Thewhole process of the proposed algorithm consists of four sub-steps:Algorithm 6.11 PageGather algorithmStep 1. Process the access log into visits.Step 2. Compute the co-occurrence frequencies between pages and create a similarity matrix.