A Novel Approach for Reducing Data Sparsity in Recommender System

Abstract

Traditionally, information sparsity is seen as a downside to user-based CF. Data sparseness is often presumed to result in a tiny amount of co-rated products or none between two consumers, leading in unreliable or unavailable similarity data, and further incurring bad recommendation quality. However, the evaluation method is often not experimentally confirmed. The impacts of information sparsity on user-based CF are evaluated in two steps in order to conduct a thorough assessment. First, weigh up the match between two users. Second, if there are similar testers, a Collaborative Filtering Prediction Technique (CFPT) is utilized to improve the sparsity issue and upgraded the nature of the suggestion. We completed a trial dependent on the MovieLens informational index, the consequences of the trial showed that the proposed methodology could successfully lessen the information sparsity issue and, as far as forecasts, our proposed methodology outflanks other essential strategies.