The collaborative filtering algorithm[1] proposed by Grouplens[2] is one of the most commonly used methods for personalized recommendation in recommendation systems [3] [4] [5] [6], and the core component of User-based collaborative filtering is the similarity measure. The traditional user similarity measurement method does not consider the influence of factors such as frequent user interest transfer and content popularity degree difference on the accuracy of the algorithm, and the existing improvement strategies cannot comprehensively consider these two factors. Based on the traditional similarity algorithm, this paper introduces influential factors such as user interest decline over time and content popularity, so as to improve the existing user similarity algorithm and to compare the actual data to prove the improved algorithm.
展开▼