Clustering-Based Multi-Criteria Recommender System Using Fuzzy C-means and Mahalanobis Distance

Over the last several years, significant progress has been achieved in recommender systems, which have been deployed in both industry and academia. Multi-Criteria recommender systems enhance accuracy, address problems like sparsity and cold start, and ensure scalability in large databases. In this paper, we try to improve the efficiency of multi-criteria recommender systems that use clustering to handle the issue of high dimensionality. We employ the fuzzy C-means clustering algorithm and construct the neighborhood set using Mahalanobis distance. We also evaluate the developed multi-criteria recommender system using a standard evaluation metric and compare its performance with other multi-criteria recommender systems that employ different clustering algorithms and similarity measures. Experimentally, we find that the recommender system using the fuzzy C-means algorithm and Mahalanobis distance exhibits better performance than other multi-criteria recommender systems on the given dataset.