Performance Comparison of Rank Aggregation Using Borda and Copeland in Recommender System

摘要

The rapid development of e-commerce will certainly be followed by an increasing number and more varied marketed products, making it confusing and time-consuming for users to choose from a large number of desired products. Consequently, a recommendation system is required to give products suggestion to the users with high accuracy. One of the most commonly techniques for the recommendation system is the collaborative filtering technique. However, this technique faces major problems, i.e., sparsity, scalability, and cold start. This paper combine clustering and ranking aggregation approaches to solve the problem. The clustering approach uses K-means algorithm. The approach is implemented into MovieLens dataset, which consists of the demographic and genre information. Meanwhile, the ranking aggregation approach uses Borda and Copeland methods to be compared. The results of the experiment show that Borda method is superior than Copeland method. The mean score of NDCG for Borda and Copeland methods are 0.6251 and 0.5649 respectively. Whilst the running time of Borda method is 63.6793 seconds faster than that of Copeland method.