Time-Weighted Uncertain Nearest Neighbor Collaborative Filtering Algorithm

Abstract

To overcome the limitations of the traditional collaborative filtering recommendation algorithm, this paper proposed a Time-Weighted Uncertain Nearest Neighbor Collaborative Filtering Algorithm (TWUNCF). According to the actual application situation of recommendation system, the author weighted the product similarity and user similarity to ensure the data validity firstly, and then calculate the similarities of user and product and choose the trusted neighbor group as the recommended object adaptively based on the weight. Experimental results show that the algorithm can be used to improve data validity according to the time attribute, and balance the impact the different groups on the recommendation result, and avoid the problems which caused by the data sparseness. Theoretical analysis and experimental demonstrations show that the algorithm this paper proposed outperforms existing algorithms in recommendation quality, and improve the system's accuracy and recommendation efficiency.