Probabilistic Association Rules for Item-Based Recommender Systems

摘要

Since the beginning of the 1990's, the Internet has constantly grown, proposing more and more services and sources of information. The challenge is no longer to provide users with data, but to improve the human/computer interactions in information systems by suggesting fair items at the right time. Modeling personal preferences enables recommender systems to identify relevant subsets of items. These systems often rely on filtering techniques based on symbolic or numerical approaches in a stochastic context. In this paper, we focus on item-based collaborative filtering (CF) techniques. We show that it may be difficult to guarantee a good accuracy for the high values of prediction when ratings are not enough shared out on the rating scale. Thus, we propose a new approach combining a classic CF algorithm with an item association model to get better predictions. We deal with this issue by exploiting probalistic skewnesses in triplets of items. We validate our model by using the MovieLens dataset and get a significant improvement as regards the High MAE measure.