The development of a parallel multi agent system for association rule mining

• Main Authors: Ahmad, Abd. Manan, Abdul Manaf, Sazali, Alias, Mohamad Ashari
• Format: Monograph
• Language: English
• Published: Faculty of Computer Science and Information System 2003
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://eprints.utm.my/id/eprint/4545/1/71868.pdf; http://eprints.utm.my/id/eprint/4545/
• Institution: Universiti Teknologi Malaysia

The continuous growth of the information on the Internet makes it necessary for users to be provided with a convenient and yet accurate tools to capture the information needed. Web usage mining has gained more popularity among researchers in discovering the users browsing behavior by mining the web server log that records all the users transactions activities. By applying it into the recommendation engine, Web personalization can be executed based on the discovered users behavior. Nevertheless, the efficiency of the generated recommendations is still an issue for reseachers. This final report focusing on the development of a usage model for predictions based on association rule and similarity measures, named ARsim. Model development will used agent technology to improve data processing time and generate recommendations rules. Aglet 1.1b3 will be agent platform for this model development. Additional parameter was used to measure the similarities between URLs, which is the time user spend on a particular page. To generate the final recommendation, similarity between URLs contained in the active user profile was calculated upon the matched Web usage profiles and finally the top-N most similar URLs are then recommended to the user. Three evaluation metrics, which is commonly used by other researchers for evaluation of Web page recommendation model, was applied to evaluate the efficacy of ARsim, namely precision, coverage and F1. Comparison to two other different techniques, traditional association rule and eVZpro found that the integration of rules and similaty measures allow only the most appropriate URLs to be recommended and thus increase the efficiency of the Web page recommendation engine.