
		<paper>
			<loc>https://jjcit.org/paper/244</loc>
			<title>A NOVEL EVIDENTIAL COLLABORATIVE FILTERING FRAMEWORK BASED ON DISCOUNTING CONFLICTING PREFERENCES</title>
			<doi>10.5455/jjcit.71-1718871610</doi>
			<authors>Khadidja Belmessous,Faouzi Sebbak,M’hamed Mataoui</authors>
			<keywords>Recommender systems,Collaborative filtering,Dempster-Shafer theory,Conflict,Fusion</keywords>
			<citation>2</citation>
			<views>2892</views>
			<downloads>582</downloads>
			<received_date>20-Jun.-2024</received_date>
			<revised_date>2-Sep.-2024 and 24-Sep.-2024</revised_date>
			<accepted_date>27-Sep.-2024</accepted_date>
			<abstract>This paper presents a novel framework to enhance Evidential Collaborative Filtering (ECF), a critical Recommender System (RS) designed for sensitive domains like healthcare and target tracking. The focus is on refining how user-rating imperfections are handled, particularly in managing conflicting preferences during neighborhood selection to boost recommendation quality. The newly proposed ECF architecture integrates a two-probabilities-focused approach with an advanced conflict-management technique, employing Deng relative entropy and the Best Worst Method. This allows for assigning more accurate reliability weights to each user, improving preference selection and rating prediction in ECF. Experimental evaluations on Movielens-100K and Flixster datasets show that our framework surpasses baselines in prediction error, precision, recall and F-score.</abstract>
		</paper>


