Selecting relevant instances for efficient and accurate collaborative filtering

摘要: Collaborative filtering uses a database about consumers' preferences to make personal product recommendations and is achieving widespread success in both E-Commerce Information Filtering Applications nowadays. However, the traditional collaborative algorithms do not scale well ever-growing number of consumers. The quality recommendation also needs be improved order gain more trust from In this paper, we present novel method improve scalability accuracy algorithm. We introduce an information theoretic approach measure relevance consumer (instance) for predicting preference given (target concept). proposed reduces training data set by selecting only highly relevant instances. Our experimental evaluation on well-known EachMovie shows that our doesn't significantly speed up prediction, but results better accuracy.