MCARS-CC: A Salable Multicontext-Aware Recommender System

作     者：Nawara, Dina Kashef, Rasha 

作者机构：Toronto Metropolitan Univ Dept Elect Comp & Biomed Engn Toronto ON M5B 2K3 Canada 

出 版 物：《IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS》 (IEEE Trans. Computat. Soc. Syst.)

年 卷 期：2024年第11卷第1期

页      面：612-624页

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

主　　题：Clustering algorithms Context modeling Scalability Partitioning algorithms Recommender systems Automobiles Collaboration Clustering consensus clustering context-aware recommender postfiltering prefiltering sentiment analysis 

摘      要：Context-aware recommendation systems (CARSs) leverage contextual information, e.g., time, location, or mood, to generate more personalized recommendations with high accuracy;however, existing CARSs fall short in: 1) handling the high sparsity of data;2) designing scalable solutions in real time;and 3) providing more personalized solutions with the current limited static contexts. This article proposes a multi-CARS based on consensus clustering (MCARS-CC) to solve these challenges. The item-based contextual information is acquired using explicit static and inferred contexts by applying sentiment analysis to the users reviews. The proposed model is experimented using contextual prefiltering and postfiltering techniques applied to two benchmark datasets, Yelp and TripAdvisor. The model is evaluated using mean absolute error (MAE), root-mean-squared error (RMSE), response time, precision, recall, and F-measure. The experimental results show that the proposed MCARS-CC model outperforms other baseline techniques using the accuracy and error-based metrics. Incorporating hypergraph partitioning algorithm (HGPA) could improve the MAE and RMSE by 25.96% and 8.94% (Yelp), respectively. Also, HGPA led to an 18.47% and 15.94% improvement ratio in terms of MAE and RMSE (TripAdvisor), respectively.