Resolving cold start and sparse data challenge in recommender systems using multi-level singular value decomposition. (September 2021)

• Record Type: Journal Article
• Authors: Vahidy Rodpysh, Keyvan; Mirabedini, Seyed Javad; Banirostam, Touraj
• Is Part Of: Computers & electrical engineering. Volume 94(2021)
• Publication Date: 2021-09
• Subjects: Recommender systems -- Singular value decomposition -- Contextual information -- Cold start -- Sparse data; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2021.107361
• Languages: English
• ISSNs: 0045-7906

Abstract

Highlights: Multi-level singluare value decomposition method with contextual information is developed to handle cold start and sparse data problem. The proposed framework towards mentinee on context features users and items in recommender system. Providing approach that applied to contextual information to recommendation. Experimenting on IMDB and STS datasets for analyze and validation. Abstract: Recommender systems estimate users' tendency and recommend a list of suitable items. Two fundamental challenges in recommender systems include cold start and sparse data. In order to overcome these challenges, employing contextual similarity measures, utilizing the features of the users and items, and applying machine learning methods have been presented. However, a method called the context feature singular value decomposition is presented as the first step. In this method, the user-context feature matrix, the item-context feature matrix, and the context similarity matrix are created to mitigate cold start. In the second step, matrices obtained in the previous step are applied as components of a multi-level singular value decomposition matrix and momentum stochastic gradient descent feature to reduce sparse data. The results obtained from the presented method are compared to those of existing approaches, indicating that the proposed method improves the accuracy of suggested entities in recommender systems Graphical abstract: Image, graphical abstract