Document Type : Original Article
Authors
1
PhD Candidate in Operations Research, Department of Management, Faculty of Administrative and Economic Sciences, Ferdowsi University of Mashhad, Mashhad, Iran. omid.valizadeh@mail.um.ac.ir
2
PhD Candidate in Operations Research, Department of Management, Faculty of Administrative and Economic Sciences, Ferdowsi University of Mashhad, Mashhad, Iran.
3
PhD Student, Université de Rennes, CNRS, CREM - UMR 6211, F-35000 Rennes, France. atieh.jafarian@univ-rennes1.fr
10.22075/mmsd.2025.38693.1015
Abstract
Background and Objectives: The capital market, as a fundamental pillar of financing and economic transparency, plays a vital role in the sustainability of firms. Despite notable advances in financial analysis, most prior studies have predominantly focused on one-dimensional evaluations of financial indicators, paying limited attention to structural heterogeneity among firms and the identification of latent patterns. Moreover, research conducted in the Iranian stock market has largely relied on traditional statistical models, which exhibit limited capacity in modeling nonlinear and complex relationships. The objective of this study is to propose an integrated framework for identifying distinct financial clusters and predicting firms’ sustainability by employing a hybrid approach that combines clustering techniques with machine learning.
Materials and Methods: Initially, a dataset comprising eight key financial sustainability indicators was collected. After preprocessing which involved outlier removal, imputing missing values using the KNN method, and standardization the data were fed into the K-Means algorithm. To determine the optimal number of clusters, the silhouette index was calculated, identifying three as the optimal solution. The resulting cluster labels were then employed as the target variable in three machine learning models (XGBoost, Random Forest, and Decision Tree), whose performance was evaluated using accuracy, precision, recall, and F1-Score metrics. Finally, through feature importance analysis, the variables contributing most significantly to cluster differentiation were identified.
Results: The clustering results revealed three distinct structural patterns among firms: a cluster characterized by high profitability, strong economic value added, and low risk; a cluster with high growth and returns but accompanied by elevated risk and cost of capital; and a cluster marked by weak profitability and limited value creation. A comparison of the machine learning models indicated that the XGBoost algorithm achieved the best predictive performance, with an accuracy of 0.97 and an F1-Score of 0.96. Furthermore, the feature importance analysis demonstrated that ROA was the most influential indicator in differentiating clusters across all three models, followed by WACC and EVA as key determinants of financial sustainability.
Conclusion: The present study demonstrates that the hybrid approach of clustering and machine learning can serve as an effective tool for identifying structural heterogeneity in the capital market. The findings suggest that managerial focus on improving ROA and reducing WACC through optimal capital structure management, along with enhancing EVA, constitutes a key strategy for strengthening firms’ financial sustainability.
Keywords
Main Subjects
)