{"title":"廓形escoreinr:通过利用局部凝聚力和最近邻分离,超越传统的网络布局","authors":"Hua-Ying Chuang , Willy Chou","doi":"10.1016/j.mex.2025.103622","DOIUrl":null,"url":null,"abstract":"<div><div>The silhouette score (SS) quantifies how well each entity fits its assigned cluster by contrasting within‐cluster cohesion with nearest other–cluster separation. Although common in other fields, SS is rarely used in bibliometrics. Using 2,252 <em>MethodsX</em> articles (2020–2024), we show how SS evaluates clustering quality in co-word networks and author collaborations, independent of the chosen algorithm. We provide R scripts to compute SS for explicit (geographic/known coordinates) and implicit (PCA/UMAP) layouts and introduce a two-axis visualization that plots publication count against SS. The framework highlights coherent clusters (high SS) and flags boundary or misassigned entities (low/negative SS) that standard network plots can obscure. This improves interpretability at term, cluster, and corpus levels and supports more defensible decisions about labels, membership, and follow-up analysis. Code is released for replication and reuse; sensitivity to distance metrics and data regimes is discussed to guide application across bibliometrics and related domains.<ul><li><span>•</span><span><div>Silhouette Scores Reveal Outliers: Silhouette scores not only validate cluster cohesion but also uncover meaningful outliers—insights often missed in traditional network layouts.</div></span></li><li><span>•</span><span><div>Novel Visualization Approach: Combining silhouette scores with publication counts enables a more nuanced visualization of co-word and collaboration networks.</div></span></li><li><span>•</span><span><div>Applied to Bibliometrics: This study applies silhouette analysis to 2252 MethodsX articles, offering new tools for evaluating clustering quality in bibliometric research.</div></span></li></ul></div></div>","PeriodicalId":18446,"journal":{"name":"MethodsX","volume":"15 ","pages":"Article 103622"},"PeriodicalIF":1.9000,"publicationDate":"2025-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"SilhouetteScoreinR: Beyond traditional network layouts by leveraging local cohesion and nearest neighbor separation\",\"authors\":\"Hua-Ying Chuang , Willy Chou\",\"doi\":\"10.1016/j.mex.2025.103622\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>The silhouette score (SS) quantifies how well each entity fits its assigned cluster by contrasting within‐cluster cohesion with nearest other–cluster separation. Although common in other fields, SS is rarely used in bibliometrics. Using 2,252 <em>MethodsX</em> articles (2020–2024), we show how SS evaluates clustering quality in co-word networks and author collaborations, independent of the chosen algorithm. We provide R scripts to compute SS for explicit (geographic/known coordinates) and implicit (PCA/UMAP) layouts and introduce a two-axis visualization that plots publication count against SS. The framework highlights coherent clusters (high SS) and flags boundary or misassigned entities (low/negative SS) that standard network plots can obscure. This improves interpretability at term, cluster, and corpus levels and supports more defensible decisions about labels, membership, and follow-up analysis. Code is released for replication and reuse; sensitivity to distance metrics and data regimes is discussed to guide application across bibliometrics and related domains.<ul><li><span>•</span><span><div>Silhouette Scores Reveal Outliers: Silhouette scores not only validate cluster cohesion but also uncover meaningful outliers—insights often missed in traditional network layouts.</div></span></li><li><span>•</span><span><div>Novel Visualization Approach: Combining silhouette scores with publication counts enables a more nuanced visualization of co-word and collaboration networks.</div></span></li><li><span>•</span><span><div>Applied to Bibliometrics: This study applies silhouette analysis to 2252 MethodsX articles, offering new tools for evaluating clustering quality in bibliometric research.</div></span></li></ul></div></div>\",\"PeriodicalId\":18446,\"journal\":{\"name\":\"MethodsX\",\"volume\":\"15 \",\"pages\":\"Article 103622\"},\"PeriodicalIF\":1.9000,\"publicationDate\":\"2025-09-11\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"MethodsX\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S2215016125004662\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"MULTIDISCIPLINARY SCIENCES\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"MethodsX","FirstCategoryId":"1085","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S2215016125004662","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"MULTIDISCIPLINARY SCIENCES","Score":null,"Total":0}
SilhouetteScoreinR: Beyond traditional network layouts by leveraging local cohesion and nearest neighbor separation
The silhouette score (SS) quantifies how well each entity fits its assigned cluster by contrasting within‐cluster cohesion with nearest other–cluster separation. Although common in other fields, SS is rarely used in bibliometrics. Using 2,252 MethodsX articles (2020–2024), we show how SS evaluates clustering quality in co-word networks and author collaborations, independent of the chosen algorithm. We provide R scripts to compute SS for explicit (geographic/known coordinates) and implicit (PCA/UMAP) layouts and introduce a two-axis visualization that plots publication count against SS. The framework highlights coherent clusters (high SS) and flags boundary or misassigned entities (low/negative SS) that standard network plots can obscure. This improves interpretability at term, cluster, and corpus levels and supports more defensible decisions about labels, membership, and follow-up analysis. Code is released for replication and reuse; sensitivity to distance metrics and data regimes is discussed to guide application across bibliometrics and related domains.
•
Silhouette Scores Reveal Outliers: Silhouette scores not only validate cluster cohesion but also uncover meaningful outliers—insights often missed in traditional network layouts.
•
Novel Visualization Approach: Combining silhouette scores with publication counts enables a more nuanced visualization of co-word and collaboration networks.
•
Applied to Bibliometrics: This study applies silhouette analysis to 2252 MethodsX articles, offering new tools for evaluating clustering quality in bibliometric research.