2015 · 109 citations · 21 references
ComputingData ScienceData MiningInformation RetrievalRecommender SystemsCollaborative FilteringEvaluation Of Recommender SystemsRecommendation EvaluationRecommendation SystemsAccuracyCorrelation-based PredictionPearson Correlation CoefficientRecommender Evaluation MetricsSimilarity ComputationSimilarity MeasuresUser Similarity Modeling
Recommender systems help users to find information that best fits their preferences and needs in an overloaded search space. Most recommender systems researches have been focused on the accuracy improvement of recommendation algorithms. Choosing appropriate similarity measure is a key to the recommender system success for this target. Pearson Correlation Coefficient (PCC) is one of the most popular similarity measures for Collaborative filtering recommender system, to evaluate how much two users are correlated. While Correlation-based prediction schemes were shown to perform well, they suffer from some limitations. In This paper we present an extension toward Pearson Correlation Coefficient measure for cases which does not exist similarity between users by using it. Experimental result on the film trust data set demonstrate via our proposed measure and PCC we can achieve better result for similarity measure than traditional PCC.
21
Paul Resnick, Neophytos Iacovou, Mitesh Suchak et al. · 1994 · 5K citations · Full text
Factorization meets the neighborhood
Yehuda Koren · 2008 · 3.9K citations
Jesús Bobadilla, Fernando Ortega, Antonio Hernando et al. · Knowledge-Based Systems · 2013 · 3.1K citations
An Algorithmic Framework for Performing Collaborative Filtering
Jonathan L. Herlocker, Joseph A. Konstan, Al Borchers et al. · ACM SIGIR Forum · 2017 · 2.6K citations