A Novel Recommendation Algorithm Considering Average Similarity and User-based Collaborative Filtering

Abstract

Received: 11 June 2019 Accepted: 18 August 2019 This paper attempts to improve the accuracy of traditional collaborative filtering recommendation algorithms. To solve the sparsity of the scoring matrix, the author designed a novel collaborative filtering recommendation algorithm based on average similarity (AS) and user-based collaborative filtering (USF). The proposed algorithm was subjected to parallelization programming on MapReduce, followed by the analysis on the parallelization of the algorithm. Next, the proposed algorithm was verified through experiments with varied ratios. The experiments show that our algorithm can compensate for the sparseness of the scoring matrix in traditional algorithms, and output accurate recommendation results. The research findings shed important new light on solving recommendation problems in the ear of big data.