Hybrid Swarm Intelligence-Based Biclustering Approach for Recommendation of Web Pages

Abstract

This chapter focuses on recommender systems based on the coherent user's browsing patterns. Biclustering approach is used to discover the aggregate usage profiles from the preprocessed Web data. A combination of Discrete Artificial Bees Colony Optimization and Simulated Annealing technique is used for optimizing the aggregate usage profiles from the preprocessed clickstream data. Web page recommendation process is structured in to two components performed online and offline with respect to Web server activity. Offline component builds the usage profiles or usage models by analyzing historical data, such as server access log file or Web logs from the server using hybrid biclustering approach. Recommendation process is the online component. Current user's session is used in the online component for capturing the user's interest so as to recommend pages to the user for next navigation. The experiment was conducted on the benchmark clickstream data (i.e. MSNBC dataset and MSWEB dataset from UCI repository). The results signify the improved prediction accuracy of recommendations using biclustering approach.