Collaborative Filtering Recommendation Model Through Active Bayesian Classifier
Gao Lin-qi, Congdong Li · 2006
Recommendation system recommends suitable products to customer through acquiring customer's requirement. The customers' classifying becomes the basis to produce recommendation. Customers' classifying has several features, such as huge sample space and class frequently changed. Traditional collaborative filtering algorithm works poor in this situation. To improve recommending quantity, a collaborative filtering model was proposed based on active Bayesian classifier. It has following features: (1) Through estimating sample's utility for classifier, a sample selecting strategy was defined to reduce the number of samples while maintaining the quality of classification. (2) The training process of classifier is a loop procedure about sampling, label and study process. The termination condition of loop may be the time restraint, suits to the on-line application to increase recommendation speed. At last, experiments ware designed at the basis of MoveLens dataset. Comparing with general collaborative filtering, the proposed algorithm has higher quality of recommendation.