<?xml version="1.0" encoding="UTF-8"?>
<ArticleSet>
  <Article>
    <Journal>
      <PublisherName>The Research Department of Economics and Management of Tadbir Nikan</PublisherName>
      <JournalTitle>Business, Marketing, and Finance Open</JournalTitle>
      <Issn></Issn>
      <Volume></Volume>
      <Issue>In Press</Issue>
      <PubDate PubStatus="epublish">
        <Year>2027</Year>
        <Month>10</Month>
        <Day>01</Day>
      </PubDate>
    </Journal>
    <ArticleTitle>Developing an Early Identification Model for Bank Bankruptcy Using Unsupervised Learning</ArticleTitle>
    <VernacularTitle>Developing an Early Identification Model for Bank Bankruptcy Using Unsupervised Learning</VernacularTitle>
    <FirstPage>1</FirstPage>
    <LastPage>20</LastPage>
    <Language>EN</Language>
    <AuthorList>
      <Author>
        <FirstName></FirstName>
        <LastName></LastName>
        <Affiliation></Affiliation>
      </Author>
      <Author>
        <FirstName></FirstName>
        <LastName></LastName>
        <Affiliation></Affiliation>
      </Author>
      <Author>
        <FirstName></FirstName>
        <LastName></LastName>
        <Affiliation></Affiliation>
      </Author>
    </AuthorList>
    <PublicationType>Journal Article</PublicationType>
    <History>
      <PubDate PubStatus="received">
        <Year>2026</Year>
        <Month>07</Month>
        <Day>28</Day>
      </PubDate>
    </History>
    <Abstract>&lt;p&gt;The present study aimed to develop a model for the early identification of bank bankruptcy using unsupervised learning. To this end, a mixed qualitative–quantitative approach was employed. In the qualitative phase, indicators associated with the early identification of bank bankruptcy were extracted through in-depth semi-structured interviews with experts and subsequently validated based on their opinions. In the quantitative phase, data relating to 13 listed banks over the period 2016–2024 were collected and analyzed, yielding 117 bank-year observations. First, canonical correlation analysis was employed to examine the relationships between the extracted indicators and bank soundness assessment criteria. Subsequently, K-means clustering was used to classify the observations into five clusters. Local anomaly detection was then applied to identify abnormal observations and banks potentially exposed to bankruptcy risk. In addition, the model results were compared with the Altman Z-score, while the performance of the proposed approach was compared with the supervised learning algorithms of support vector machine and decision tree. The findings showed that, out of 117 observations, 92 observations, equivalent to 78.6%, were identified as anomalous. Of these, 42.7% were classified within the cluster of financially unhealthy banks or banks exposed to bankruptcy risk, whereas 57.3% were placed within the cluster of financially healthy banks. According to the Altman Z-score criterion, 13 observations, equivalent to 11.1%, were classified as financially healthy, whereas 104 observations, equivalent to 88.9%, were classified as financially distressed. Comparative results indicated that the proposed model achieved reported detection rates of 92.3% and 90.4% for healthy and unhealthy observations, respectively, and outperformed the support vector machine and decision tree algorithms in distinguishing financially healthy banks from banks exposed to bankruptcy risk. Overall, the findings indicate that integrating expert-based indicators with clustering and anomaly-detection techniques can provide a practical approach for the early identification of abnormal financial conditions in banks and can contribute to strengthening bank financial soundness monitoring systems.&lt;/p&gt;</Abstract>
    <ObjectList>
      <Object Type="keyword">
        <Param Name="value">bank bankruptcy</Param>
      </Object>
      <Object Type="keyword">
        <Param Name="value">early identification</Param>
      </Object>
      <Object Type="keyword">
        <Param Name="value">unsupervised learning</Param>
      </Object>
      <Object Type="keyword">
        <Param Name="value">anomaly detection</Param>
      </Object>
      <Object Type="keyword">
        <Param Name="value">clustering</Param>
      </Object>
    </ObjectList>
    <ArchiveCopySource DocType="pdf">https://bmfopen.com/index.php/bmfopen/article/download/707/516</ArchiveCopySource>
  </Article>
</ArticleSet>
