Association-rules-based recommender system for personalization in adaptive web-based applications

Abstract

Personalization systems based upon the analysis of users' surfing behavior imply three phases: data collection, pattern discovery and recommendation. Due to the dimension of log files and high processing time, the first two phases are achieved offline, in a batch process. In this article, we propose Wise Recommender System (WRS), an architecture for adaptive web applications. Within this framework, usage data is implicitly obtained by the data collection submodule. This allows for the extraction of usage data, online and in real time, by using a proactive approach. For the pattern discovery, we efficiently used association rule mining among both frequent and infrequent items. This is due to the fact that the pattern discovery module transactionally processes users' sessions and uses incremental storage of rules. Finally, we will show that WRS can be easily implemented within any web application, thanks to the efficient integration of the three phases into an online transactional process. © 2010 Springer-Verlag.