{"title":"Degree-Related Bias in Link Prediction","authors":"Yu Wang, Tyler Derr","doi":"10.1109/ICDMW58026.2022.00103","DOIUrl":null,"url":null,"abstract":"Link prediction is a fundamental problem for network-structured data and has achieved unprecedented success in many real-world applications. Despite the significant progress being made towards improving its performance by characterizing underlined topological patterns or leveraging representation learning, few works have focused on the imbalanced performance among nodes of different degrees. In this paper, we propose a novel problem, degree-related bias and evaluation bias, on link prediction with an emphasis on recommender system applications. We first empirically demonstrate the performance differ-ence among nodes with different degrees and then theoretically prove that Recall is an unbiased evaluation metric compared with Fl, NDCG and Precision. Furthermore, we show that under the unbiased evaluation metric Recall, low-degree nodes tend to have higher performance than high-degree nodes in link prediction.","PeriodicalId":146687,"journal":{"name":"2022 IEEE International Conference on Data Mining Workshops (ICDMW)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2022-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 IEEE International Conference on Data Mining Workshops (ICDMW)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICDMW58026.2022.00103","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 2
Abstract
Link prediction is a fundamental problem for network-structured data and has achieved unprecedented success in many real-world applications. Despite the significant progress being made towards improving its performance by characterizing underlined topological patterns or leveraging representation learning, few works have focused on the imbalanced performance among nodes of different degrees. In this paper, we propose a novel problem, degree-related bias and evaluation bias, on link prediction with an emphasis on recommender system applications. We first empirically demonstrate the performance differ-ence among nodes with different degrees and then theoretically prove that Recall is an unbiased evaluation metric compared with Fl, NDCG and Precision. Furthermore, we show that under the unbiased evaluation metric Recall, low-degree nodes tend to have higher performance than high-degree nodes in link prediction.