Novel Boosting Frameworks to Improve the Performance of Collaborative Filtering

Recommender systems are often based on collaborative ltering. Previous researches on collaborative ltering mainly focus on one single recommender or formulating hybrid with dierent approaches. In consideration of the problems of sparsity, recommender error rate, sample weight update, and potential, we adapt AdaBoost and propose two novel boosting frameworks for collaborative ltering. Each of the frameworks combines multiple homogeneous recommenders, which are based on the same collaborative ltering algorithm with dierent sample weights. We use seven popular collaborative ltering algorithms to evaluate the two frameworks with two MovieLens datasets of dierent scale. Experimental result shows the proposed frameworks improve the performance of collaborative ltering.