Improving news articles recommendations via user clustering.

Although commonly only item clustering is suggested by Web mining techniques for news articles recommendation systems, one of the various tasks of personalized recommendation is categorization of Web users. With the rapid explosion of online news articles, predicting user-browsing behavior using collaborative filtering (CF) techniques has gained much attention in the web personalization area. However common CF techniques suffer from problems like low accuracy and performance. This research proposes a new personalized recommendation approach that integrates both user and text clustering based on our developed algorithm, W-kmeans, with other information retrieval (IR) techniques, like text categorization and summarization in order to provide users with the articles that match their profiles. Our system can easily adapt over time to divertive user preferences. Furthermore, experimental results show that by aggregating item and user clustering with multiple IR techniques like categorization and summarization, our recommender generates results that outperform the cases where each or both of them are used, but clustering is not applied. [ABSTRACT FROM AUTHOR]