An Improved Collaborative Filtering Based on a Weighted Network and Triadic Closure
Wangpeng Zhan, Qing Li · 2015
Collaborative Filtering (CF) is a successful technique used by the personalized recommendation system. The core of CF is the metric of similarity between two users, which usually uses the Pearson correlation metric. However, traditional similarity metrics have a low accuracy as a result of data sparseness. In this paper, we present a new metric which is based on a weighted network and triadic closure to improve the accuracy of similarity. The weighted network is built on each pair of users and the weight is the number of common rating items. The triadic closure is used to calculate the connection intensity between two users in the weighted network. The experimental results show that the new similarity metric is effective to improve the recommender results.