Improving the performance of collaborative filtering recommender systems using deep features extraction

The growing volume of information on the Web and the Internet has made difficult the process of deciding and selecting the information, data, or products needed for many web users. This study proposed a novel method to improve the performance of collaborative filtering recommender systems. The goal is to use deep feature extraction to provide more effective and desirable recommendations to the user. In the preprocessing step, the input data is first entered into the primary processing system and the feature values are normalized. Then, in order to perform more accurate calculations and reduce computation time, the size of the data is reduced using a deep belief network (DBN), while extracting deep features. Afterwards, using the collaborative filtering technique, the recommended items are offerd to the user. Finally, according to the system outputs in the recommendation to the user, the accuracy of the proposed items is evaluated. To evaluate the proposed method, a comparison of its performance on the real-world MovieLens dataset with basic methods has been used. Experimental results showed that the proposed method has better performance in terms of coverage and support than other compared methods.