Personalized Book Recommendation Method Based on the Improved Similarity

With the rapid development of network technology, online reading platforms have emerged one after another, providing a new opportunity for book recommendation. Collaborative filtering makes use of the preferences of neighbor users to recommend and predict the interests of target users, in which similarity calculation is the key point. Because the traditional similarity methods cannot take full benefit of the potential relationship between readers. As the result of data sparsity, the similarity matrix is too sparse, which ultimately leads to low recommendation accuracy. This article quantifies the reader's author preferences to build a readers' similarity formula incorporating author preference and Pearson coefficient by introducing Jaccard coefficient, in order to describe the association between readers more comprehensively. According to the proposed similarity, the improved collaborative filtering algorithm can improve the quality of book recommendations significantly. Finally, this paper performs simulation experiments on the Book-Crossing dataset. The results show that the collaborative filtering algorithm based on the improved similarity can effectively improve the quality of personalized book recommendation.