群集演算法於財務風險等級評估之應用──以大專院校公開財報資料為例

Applying Clustering Algorithms with Financial Risk Ranking for Universities and Colleges

受到本國人口結構轉型與全球高等教育競爭加劇影響，國內私立大專校院因生源不足陷入財務困境已引起社會各界高度重視。如何對私立大專校院財務加以評價，並予以劃分出風險等級更是目前大家所關心的焦點。然而，我國針對私立大專校院財務風險研究的證據仍不多。為此，本研究以非監督式學習篩選出代表性財務指標，並利用K-Medoids演算法為基礎建立學校分群，藉以提供私立大專校院作為風險等級評分的依據。分析結果發現，本文所篩選出的八個財務指標具有財務風險評鑑辨識度；且經實際比較發現，透過群集分組能有效劃分風險等級，提供管理者一個便利且有效的風險管理方式。

Affected by the transformation of the country's population structure and the intensified competition in global higher education, domestic private colleges and universities have fallen into financial distress due to insufficient student resources, which have attracted great attention from all walks of life. The focus of everyone's concern at present is how to evaluate the finances of private colleges and universities and classify the risk level. However, there is still little evidence on the financial risk research of private colleges and universities in our country. To this end, this study uses unsupervised learning to screen out representative financial indicators and uses the K-Medoids algorithm as the basis to establish school clusters, so as to provide private colleges and universities with the basis for risk level scoring. The results of the analysis show that the eight financial indicators that were picked out for this paper can be used for financial evaluation. Also, a real-world comparison shows that the risk level can be effectively divided by cluster grouping, which gives managers an easy and effective way to handle risk.