本研究探討基於機器學習技術的研究者關聯性分析方法，並以輔仁大學機構典藏中的期刊論文資料為案例，進行研究者間的關聯挖掘與視覺化呈現。研究結合自然語言處理（Natural Language Processing，NLP）技術，對文獻資料進行特徵提取與向量化，並運用三種分群演算法（階層式分群、K-means、DBSCAN）進行比較分析。結果顯示，K-means 分群效果最佳，於指定資料群集中透過觀察題名及使用TF-IDF方法計算群集關鍵詞，發現其能在無關資料占比較低的情況下，有效聚集相關文章，並能呈現研究者的主要研究領域與次要專業情形。最終，參考社會網絡分析技術將研究者關聯性以視覺化圖形呈現，進一步促進研究者之間的合作與跨領域交流。本研究不僅提升了機構典藏系統的應用價值，亦為研究者關聯性分析提供了一套具體的實施方案。

This study uses journal articles from Fu Jen Catholic University’s institutional repository as a case study to explore researcher relationship analysis methods based on machine learning techniques. Our methods involve extracting features and vectorizing textual data by integrating natural language processing (NLP) techniques, and utilizing three clustering algorithms—hierarchical clustering, K-means, and DBSCAN—for comparative analysis. The results indicate that K-means achieves the best clustering performance. Through title analysis and calculating cluster keywords using the TF-IDF method, K-means effectively groups related articles with minimal irrelevant data and reveals researchers’ primary and secondary expertise. Finally, the study visualizes researcher relationships by social network analysis and promotes collaboration and interdisciplinary exchange among researchers. This research enhances the practical value of institutional repositories and provides a concrete implementation framework for researcher relationship analysis.

機構典藏，自然語言處理，文本分群，研究者關聯性分析

Institutional repository, Natural language Processing, Text clustering, Researcher relationship analysis