Collaborative Filtering-Based Music Recommendation in Spark Architecture

Abstract

The use of recommendation algorithms to recommend music MOOC resources is a method that is gradually gaining ground in people’s lives along with the development of the Internet. The often used ALS collaborative filtering algorithm has an irreplaceable role in personalised recommender systems via the Spark MLlib platform. In the study, it is investigated how Spark can be used to implement efficient music system recommendations. The collaborative filtering algorithm based on the ALS model in the Spark architecture is currently the most widely used technique in recommendation algorithms, allowing for the analysis and optimisation of computational techniques. The project-based collaborative filtering algorithm used in the article enables the recommendation of music by avoiding personal information about the user. More accurate user recommendations are achieved by predicting the user’s preferences and focusing on the top ranked and highly preferred music recommendations. The method improves the performance of the recommendation algorithm, which is optimised by Spark shuffle on top of resource optimisation, and its performance improved by 54.8% after optimisation compared to when there is no optimisation.