Web Recommendation Framework based on Association Rules Coverage to be Applied for Site Modification

This paper introduces a Web Recommendation Framework based on the usage history to be applied for Site Modification as one of the applications of Web Usage Mining (WUM) that is applicable for online business and marketing applications. The framework focuses on the three main interdependent tasks for performing WUM which are Preprocessing, Pattern Discovery and Pattern Analysis. In Preprocessing, we remove all irrelevant users ’ requests from the web server log file leading to log reduction followed by users ’ identification then sessions ’ identification. We take into consideration the ambiguity found in some researches concerning the HTTP common methods. In Pattern Discovery, we extract Association Rules that can be used to relate pages that are most often referenced together in a single server session and may not be directly connected to one another via hyperlinks. In Pattern Analysis, we analyze the set of generated rules independent of the website’s topology to extract valid set of rules that achieves highest coverage for the dataset. Our experimental results confirmed that calculating association rules coverage in our case study can lead to the best rules to be provided as recommendations that can help Web designers to restructure their Web site, Web applications or even portals to better serve Web customers