Adaptive method to improve web recommendation system for anonymous users

Today the major concerns are not the availability of information but rather obtaining the right information. Web Recommendation system is a specific type of Web personalization system technique that attempts to predict the user next browsing activity then recommend the web pages items that are likely to be of interest to the user. The ability of predicting the next visited pages and recommending it to the short term navigation user (anonymous user) is highly needed. Presently, there are many recommendation systems (e.g. Analog, Web Miner, WebPersonalizer, PACT,SWARS, EntreeC, SUGGEST, one-and-only items, Hybrid and NEWER) that can be used to make recommendation to the current online user, but, recommendation to anonymous users needs to be adaptive (up to date) to the changes in users‟ interests‟ over time. This research focuses on improving the prediction of the next visited web pages and introduces them to current anonymous user. An enhanced classification algorithm is used to assign the current anonymous user to the best web navigation profile. As the users‟ interests change over time, the recommender system has the ability to modify the current web navigation profiles and keep them updated. These adaptive profiles help the prediction engine to predict and then recommend the next visited pages to the current user in an accurate manner. This research proposed two web page recommendation systems. The first is iPACT, an improved recommendation system based on PACT methodology to demonstrate the prediction accuracy of the proposed enhanced classification algorithm in this research. The prediction accuracy was evaluated against two previous recommendation systems PACT and HyperGraph. The second is Adaptive Web page Recommendation System (AWRS) which combines the classification algorithm of iPACT in addition to the ability of adaptive recommending due to the changes of the users‟ interests and weighting methods to deal with unvisited or new added pages. For the evaluating purpose, the experiments were carried out on the public CTI logs file dataset which contains the preprocessed and filtered sessionized data for the main DePaul CTI Web server. AWRS was evaluated and shows better performance as compared to several recommendation systems namely, iPACT, Association Rules and Hybrid systems. Based on the experimental results, the outcome of a considerably good accuracy is mainly due to the right classification of the current user to the best web navigation profile with similar browsing activities. Also, the adaptive phase is able to update the web navigation profile(s) based on the interest‟s changes and predict the next visited pages in accurate manner to the anonymous users based on their early stage navigation. This research opens a wide range of future works to be considered,including the investigation of the dependency between the recommended web pages for each web navigation profile, investigating the quality of the method on different datasets, and finally, the possibility to apply the proposed method in other area like the misuse detection systems