Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Open Access

Peer-reviewed

Research Article

The application of structural and machine learning models to predict the default risk of listed companies in the Iranian capital market

  • Pejman Peykani ,

    Roles Conceptualization, Investigation, Methodology, Project administration, Software, Supervision, Validation, Writing – review & editing

    pejman.peykani@yahoo.com

    Affiliation School of Industrial Engineering, Iran University of Science and Technology, Tehran, Iran

    ⨯
  • Mostafa Sargolzaei,

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Software, Supervision, Validation, Visualization, Writing – review & editing

    Affiliation Department of Finance and Banking, Faculty of Management and Accounting, Allameh Tabataba’i University, Tehran, Iran

    ⨯
  • Negin Sanadgol,

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Visualization, Writing – original draft, Writing – review & editing

    Affiliation School of Management, Economics and Progress Engineering, Iran University of Science and Technology, Tehran, Iran

    ⨯
  • Amir Takaloo,

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Department of Finance and Banking, Faculty of Management and Accounting, Allameh Tabataba’i University, Tehran, Iran

    ⨯
  • Hamidreza Kamyabfar

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Resources, Software, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Department of Finance and Banking, Faculty of Management and Accounting, Allameh Tabataba’i University, Tehran, Iran

    ⨯

Abstract

Inattention of economic policymakers to default risk and making inappropriate decisions related to this risk in the banking system and financial institutions can have many economic, political and social consequences. In this research, it has been tried to calculate the default risk of companies listed in the capital market of Iran. To achieve this goal, two structural models of Merton and Geske, two machine learning models of Random Forest and Gradient Boosted Decision Tree, as well as financial information of companies listed in the Iranian capital market during the years 2016 to 2021 have been used. Another goal of this research is to measure the predictive power of the four models presented in the calculation of default risk. The results obtained from the calculation of the default rate of the investigated companies show that 50 companies listed in the Iranian capital market (46 different companies) have defaulted during the 5-year research period and are subject to the Bankruptcy Article of the Iranian Trade Law. Also, the results obtained from the ROC curves for the predictive power of the presented models show that the structural models of Merton and Geske have almost equal power, but the predictive power of the Random Forest model is a little more than the Gradient Boosted Decision Tree model.

Citation: Peykani P, Sargolzaei M, Sanadgol N, Takaloo A, Kamyabfar H (2023) The application of structural and machine learning models to predict the default risk of listed companies in the Iranian capital market. PLoS ONE 18(11): e0292081. https://doi.org/10.1371/journal.pone.0292081

Editor: Muhammad Fareed, Universiti Utara Malaysia, MALAYSIA

Received: April 30, 2023; Accepted: September 12, 2023; Published: November 27, 2023

Copyright: © 2023 Peykani et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are within the paper.

Funding: The authors received no specific funding for this work.

Competing interests: NO authors have competing interests The authors have declared that no competing interests exist.

1. Introduction

Identifying types of risk, measuring them accurately and trying to minimize them is one of the most important goals of financial sciences worldwide [1–34]. One of the most important risks that has been in semi-traditional and modern human societies for a long time is default risk, or in more general terms, credit risk with default and bankruptcy [35–71]. The bankruptcy of financial institutions and various industries during the great credit crisis of 2008 is a sign of the importance of default risk and credit risk in the world economy [44, 47, 72–82]. In Iran, the unprofitability of some companies, which ultimately leads to their default and bankruptcy (being subject to Bankruptcy Article), has been one of the old problems of the financial markets.

Another problem that default risk creates for the financial markets, especially the debt market, is that in order to avoid the repetition of problems, strong guarantees are considered for the applicants for issuing debt securities. Finally, most of these guarantees cause problems for the government and government-owned banks [51]. These guarantees practically make the issuance of debt securities impossible or expensive for many companies. On the other hand, it creates obligations on the banking system that are not fundamentally related to it [83–86]. Credit rating is considered as an alternative method for guaranteeing securities; But this method also has its associated costs and is mostly reserved for large-sized companies. Increasing the ability of financial institutions, including banks, institutional investors such as investment companies, holding companies, and investment funds, in estimating credit risk is one of the necessities for the growth and maturity of the debt market. This issue causes these financial institutions to allocate loans to companies with a better view and face less risk of bankruptcy of debtors and related costs, including financial, economic and social costs [87–98].

It is very important to mention that one of the missing links of the default risk is to try to get a correct understanding and picture of the credit risk of each applicant for loans and debt securities. Obviously, with such a possibility, the lending institution can have a more accurate estimate of the applicant, and a more active primary and secondary market for debt securities will be created, in which debt securities are valued according to the risk and the investor invests in these securities according to the risk and expected return [99–104].

We know that many researches have been done on credit risk. In this research, an attempt has been made to examine default risk. Until now, research has been based on models where one of their simplifying assumptions is that the debt is coupon-free. In this research, Geske model is used, which has removed this assumption and the company can have coupon debt. Also, due to the increasing popularity of machine learning models, two machine learning techniques called Forest and Gradient Boosted Decision Tree have also been used to estimate the default risk of companies. Random Forest and Gradient Boosted Decision Tree are used for the first time to estimate the default risk of Iranian capital market companies. In this research, it has been tried to predict the default risk of companies in Iran’s capital market by using the four proposed models. Also, another achievement of this research will be to measure the predictive power of four default risk measurement models that can be used in financial institutions, including banks, financing companies, rating companies, and investment companies.

The rest of this article is organized as follows. In section 2, researches that have been done on this issue in the past are mentioned. In section 3, all the equations of the proposed models are stated, which were necessary to achieve the goal of this article. Also, in section 4, the results obtained from the proposed models are analyzed based on the information used. Finally, in section 5, the final conclusions are stated and the suggestions for the future to expand the concept are introduced.

2. Literature review

In this section, an overview of applied research in the field of default risk prediction is presented, which examines the default of different financial institutions using different models. We have tried to show the research done on structural models and models based on learning machine in Table 1.

thumbnail
Download:
Table 1. Conducted researches on structural models and machine learning.

https://doi.org/10.1371/journal.pone.0292081.t001

As can be seen in Table 1, it is the first time that a combination of structural models and machine learning algorithms have been investigated using the financial information of companies in Iran’s capital market.

3. Methodology

In this section, we intend to express the equations used in the proposed models and the different conditions of the models. In this research, two classical structural models including Merton and Geske models and two models based on machine learning methods including models based on the Random Forest method and the Gradient Boosted Decision Tree method are used. In this article, we have tried to compare structural models and machine learning algorithms for the first time about the Iranian capital market. The aim of the research is to achieve the best estimate of the default risk of capital market companies in Iran, and to achieve this goal, the financial information of all capital market companies during the years 2016 to 2021 has been used.

In the first stage, we prepare a list of all capital market companies in the years 2016 to 2021. Then, among these companies, the companies that have been subject to Bankruptcy Article of the Iranian Trade Law during this period are identified. According to Bankruptcy Article of the Iranian Trade Law, if at least half of a company’s capital is lost, the board of directors must immediately invite the shareholders to an extraordinary general meeting to discuss the issue of liquidation or survival of the company.

3.1 Merton and Geske models

The Black-Scholes pricing equation is as follows: (1)

In Eq 1, s0, V, T and r are respectively the value of the company’s shares, the value of the company’s assets, the maturity of the debt and the risk-free interest rate. In addition, in the Black-Scholes equation, σv means the price fluctuations of the Underlying Asset (the company’s assets), which is considered as the fluctuation of the company’s asset value in the Merton model.

In Geske model, it is assumed that the company has two types of long-term and short-term debt. The nominal value of the company’s short-term debt is called M1 and the nominal value of the company’s long-term debt is called M2. It is also assumed that the maturity of the company’s short-term debt is T1 and the maturity of the long-term debt is T2, with the assumption that T2 is greater than T1. Also, we assume that the company’s debts have a coupon in the middle of the year. At the time of coupon payment (short-term debt), if the value of the company’s assets is equal to the nominal value of the short-term debt and the value of the long-term debt, the shareholders will pay the debt, otherwise the company will default. As a result, we can consider the minimum value of the company to not default at the time of payment of short-term debt or , equal to Eq 2, where is equal to the value of long-term debt at T1.

(2)

Eq 2 becomes Eq 3 considering that in any company, the value of debt is equal to assets minus equity.

(3)

According to the Geske model, the company’s stock value at time t is the following equation, and it is worth mentioning that N() function is a bivariate cumulative normal distribution function: (4)

Also, K1 and K2 are expressed in Eq 5.

(5)

In Geske model, it is assumed that companies pay off their debt at the maturity of the coupon by issuing new shares for the benefit of creditors. The value of is the amount that if the value of the company is lower at the time of coupon payment, the company will not be able to issue new shares. Also, the probability of default in T1 or T2, the probability of default in T1 and the future probability of default in T2 if there is no default in T1 are shown in Eqs 6 to 8, respectively.

(6)(7)(8)

Finally, the required variables in the proposed models are presented in the Table 2.

thumbnail
Download:
Table 2. Variables used in structural models.

https://doi.org/10.1371/journal.pone.0292081.t002

In the following, the method of estimating the variables of the models is explained. The information related to the stock price of the companies as a determinant of the equity value has been extracted through Tseclient software. The book value of companies’ debts is obtained through Codal site and the risk-free interest rate is obtained through the Central Bank of Iran site. Also, the maturity of the debt is equal to 1 year. In Geske model, short-term debt and long-term debt are equal to the company’s current liabilities and non-current liabilities. The maturity date of short-term debt is 6 months and long-term debt is 1 year. It is worth mentioning that to determine , σv and V, by adding the famous equation of the relationship between the underlying asset price and the option price, it is done as follows: (9)

By adding Eq 9 and based on a numerical algorithm, and solving the proposed equations, the unknown parameters are obtained. To calculate the parameters, we first give an initial value to V and σv. Then we set the initial value of the asset value equal to the sum of the market value of the company’s stock and the nominal value of its debt, and calculate the initial value of the fluctuation of the asset value by setting ∂S/∂V and the fluctuation value of the company’s stock price equal to 1. Then, using the obtained values, we calculate d1 and d2, and the model provides values of S and σs. Considering the real values of S and σs and the values obtained from the model, it is tried to minimize Eq 10 by changing the value of V and σv.

(10)

The above equation for Merton’s model can be solved through Excel software and using the solver plugin. For Geske model, it is done from Python software and Scipy library and by solving two Eqs 4 and 9.

3.2 Machine learning models

In models based on machine learning algorithms, data needs to be labeled. First, a list of companies that are subject to Bankruptcy Article in the years 2016 to 2021 is prepared and labeled as defaulter companies. In Table 3, the symbol of the companies that subject to Bankruptcy Article in the years 2016 to 2021 is presented. In total, the data of 44 different companies subject to Bankruptcy Article has been presented, and then according to the mentioned procedure, the information of one year of healthy companies is also examined. It should be noted that the data set is divided into two parts based on the priority of the financial years and the training and test data sets are created. As a result, the data of the financial years leading to 2016 to 2019 were used as training data and the information of the companies in the financial years leading to 2020 and 2021 were also used as the test data set, which includes 9 companies subject to Bankruptcy Article and 91 healthy companies. The rest of the information, including the information of 206 companies, is used as a training dataset. It is important to mention that no changes are applied to the data and the data is used in raw form. In the model based on the Random Forest method, first the dataset is divided into a large number of random subsets. Then a tree is drawn for each subset. It should be noted that the explanatory variables of each tree are also selected based on a number of variables randomly selected from among all variables. Finally, each data is selected using the majority of labels given to it by different trees [124, 132].

thumbnail
Download:
Table 3. Companies subject to Bankruptcy Article during 2016 to 2021.

https://doi.org/10.1371/journal.pone.0292081.t003

The data related to the information of the test data set companies are used to estimate the default of these companies. In the model based on the Gradient Boosted Decision Tree model and unlike the Random Forest model which is based on multiple decision trees, a tree is used to classify assets. In this model, an effort is made to focus on the error of the tree and by using this way, after each classification stage, more weight is given to the samples whose classification was wrong, and the error is reduced. Continue this method until the error reaches zero and the basic method of Gradient Boosted Decision Tree is used.

In the model based on Random Forest and Gradient-Boosted Decision Trees, the information of defaulter companies is used in each year along with number of variables of that company. The used variables can be a wide range of variables based on the registered and legal information of the company, such as the type of company, shareholders, date and place of registration, subject of activity, etc., to variables based on financial statements, such as types of financial ratios and gross and operating profit margins. or include variables based on the company’s activity and operations, such as the type of product or service produced or provided, or information related to the company’s stock price.

Since our field of study is companies whose shares are traded on the capital market, it has been tried to use variables in two categories, variables based on financial statements and variables based on information related to the company’s stock price, as the most basic information available to the company in the capital market.

According to the aforementioned, the variables of the training and test data sets are displayed in Table 4 [124].

thumbnail
Download:
Table 4. Variables used in machine learning algorithms.

https://doi.org/10.1371/journal.pone.0292081.t004

In this research, the ROC parameter is used to check the predictive power of the probabilities provided by each model and the parametric t and Wilcoxon tests to measure the existence of differences between the models. By using four indices of confusion matrix, several indices are calculated to estimate the strength of the model from different dimensions. How to calculate some of these indicators is shown in Table 5.

thumbnail
Download:
Table 5. Indexes for measuring classification models.

https://doi.org/10.1371/journal.pone.0292081.t005

True Positive (TP) is data that is correctly detected as positive. For example, in a default detection model of a company, if the healthiness of the company is considered positive, a company that is healthy in the real world will be correctly classified by the healthy model. TP is known as the correct identification of the model. Also, False Positive (FP) is data that is mistakenly considered positive. In the previous example it is equal to a defaulter company that is mistakenly recognized as a healthy company. FP is equivalent to Type I Error in the statistical hypothesis test. Next, it should be mentioned that True Negative (TN) is data that is correctly recognized as negative, such as a defaulter company whose default is correctly predicted. Finally, False Negative (FN) is data that is wrongly detected as negative, meaning a healthy company that is wrongly classified as a defaulter. FN is equivalent to Type II Error in statistical hypothesis test.

ROC is a parameter by which the ability of the model is measured using TPR and FPR indices. The TPR index or model sensitivity measure calculates the ratio of detected TPs by the model to all positive samples. The FPR index also calculates the model error rate using the ratio of detected FPs to total negative samples. In continuous classification models, a threshold type is defined on the continuous output as a criterion for the final classification of the data. For example, in the default risk measurement model of the investigated companies, the output for each company is obtained in the form of a probability of default. Considering a probability as a threshold, companies are divided into two healthy or defaulter categories. By changing the threshold value, different values of TPR and FPR are obtained. If the threshold value is considered high, defaulter firms will be less likely to be classified as healthy, which means the FPR will be low. On the other hand, it is possible that a number of healthy companies will be classified as defaulters, which will reduce the TPR. If the threshold value is considered low, it is more likely that the defaulter companies will be identified as healthy companies and the FPR value will be higher. The ROC curve is drawn in a two-dimensional space where the vertical column is TPR and the horizontal column is FPR, based on different thresholds. Finally, for each threshold, there will be a different value of TPR and FDR. The most favorable point for a model will be the point (1, 0) in the northwest of the figure. The line drawn in the figure and passing through the points (0, 0), (0.5, 0.5) and (1, 1) shows the performance of a random classification model. It is worth noting that the model with a larger surface area under the ROC figure is a more desirable model. The surface area under the figure is a number between 0 and 1.

In order to compare each pair of models in this research, the data generated by the models are analyzed as ordered pairs (each pair consisting of the possibilities of a company). Also, the T- Statistic is obtained based on Eqs 11 and 12.

(11)(12)

In the above equations, the variables XA, XB, nA, nB and S2 are respectively the sample of the first model, the sample of the second model, the number of samples of the first model, the number of samples of the second model and the mixed variance of the data of both models. In this test, the hypothesis H0 is equal to the equality of the average of the two samples and the hypothesis H1 is equal to the rejection of the hypothesis H_0 and as a result the inequality of the average of the two samples. Like the T-test, the Wilcoxon test is designed to compare the average of two populations, with the difference that in this test, the assumption of normality of the statistical population is not important.

For Wilcoxon test for n data by each model (2n in total), we need to go through 5 steps. In the first step, the absolute value of the difference of both pairs of data is |XAi − XBi| is calculated. Then, for the difference of both pairs, the value of the sgn function is obtained, which is defined as 1 if the difference is positive (XAi > XBi) and -1 if it is negative, and 0 if it is 0. In the second step, zero is removed from the values obtained from the sgn function. In the third step, the values obtained from Absolute function are arranged in increasing order. In the fourth step, the sorted Absolute values are ranked, with the lowest rank being given 1. If the two Absolute function values are equal, the average ranks are assigned to them. The rank assigned to the output of the Absolute function of each pair (i) is denoted as Ri. Finally, w statistic is determined according to Eq 13.

(13)

The distribution of w statistic is equal to zero and its variance is obtained from Eq 14.

(14)

In the Wilcoxon test, the statistical assumptions are the same as the T-test, and finally, according to the obtained value of w and comparing it with the critical value, the hypothesis H0 is confirmed or rejected.

4. Results

In this section, findings resulting from the implementation of structural models and machine learning algorithms are presented. Also, the point that should be mentioned here is that the complete results obtained from the models examined in this research are displayed in the Appendices A to D in S1 File.

Merton model has been implemented using Excel software, while Geske model and models based on Random Forest and Gradient Boosted Decision Tree have been analyzed using Python software. First, a table of the frequency of listed companies subject to Bankruptcy Article based on different industries in Iran’s capital market is presented in Table 6.

thumbnail
Download:
Table 6. Companies subject to Bankruptcy Article by industry.

https://doi.org/10.1371/journal.pone.0292081.t006

From the point of view of the number of companies included in each year, it seems that this trend has decreased during the years 2016 to 2021, which is evident in Table 7.

thumbnail
Download:
Table 7. Number of companies subject to Bankruptcy Article in 2016 to 2021.

https://doi.org/10.1371/journal.pone.0292081.t007

Now, the statistical indicators of each feature including average, standard deviation, minimum, first quartile, second quartile, third quartile and maximum are presented in three formats of all companies, defaulted companies and healthy companies respectively in Tables 8–10.

thumbnail
Download:
Table 8. Statistical indicators of data set characteristics—All companies.

https://doi.org/10.1371/journal.pone.0292081.t008

thumbnail
Download:
Table 9. Statistical indicators of data set characteristics–Defaulted companies.

https://doi.org/10.1371/journal.pone.0292081.t009

thumbnail
Download:
Table 10. Statistical indicators of data set characteristics—Healthy companies.

https://doi.org/10.1371/journal.pone.0292081.t010

After implementing the models in different situations, we have investigated the capabilities of each model using the ROC. The entire dataset used includes 306 companies (50 companies subject to Bankruptcy Article and 256 healthy companies). It should be noted that the information of 206 companies, including 40 companies subject to Bankruptcy Article and 166 healthy companies, was used as training data between 2016 and 2019. Also, 100 companies, including 9 companies subject to Bankruptcy Article and 91 healthy companies between 2020 and 2021, have been used as test data. Machine learning algorithms have been implemented on all data sets through the use of training data and test data. Then the structural models have been implemented on the data of 2020 and 2021 and finally the performance of the used models have been compared using ROC. In Fig 1, the Roc curve for all four models is displayed. A model with a higher area under the curve has better performance. Considering that the number of the surface area under the ROC curve is between 0 and 1, if this number is closer to 1, the model has more power.

thumbnail
Download:
Fig 1. ROC curves for four models.

https://doi.org/10.1371/journal.pone.0292081.g001

According to the numerical results presented in Table 11, the model based on the Random Forest method with the ROC curve area of 0.97 has the highest power to detect defaulted companies. Then, the model based on the Gradient Boosted Decision Tree method, with the area under the ROC curve of 0.91, ranks next. The numbers related to structural models have a significant gap with models based on machine learning techniques, so that the score of Geske model is 0.58 and Merton model is recorded with a lower number of 0.53, and they are in the next ranks from the point of view of recognition power. It should also be noted that the models have been compared two by two using the Wilcoxon model and the T-test, and the results of the statistical tests are displayed on the left side of the results table.

thumbnail
Download:
Table 11. Research results.

https://doi.org/10.1371/journal.pone.0292081.t011

Among the performance comparison tests of the models, only the similarity of the performance of the two models based on Random Forest and Gradient Boosted Decision Tree has been confirmed based on the t-test statistic. Models based on Random Forest and Gradient Boosted Decision Tree have closer performance. In Table 12, some indicators of measuring the power of two models are presented.

thumbnail
Download:
Table 12. Comparative indices of machine learning models.

https://doi.org/10.1371/journal.pone.0292081.t012

Among the indicators presented in Table 12 and the AUC number, the AUC number is the most important. After the AUC number, a parameter that is important in financial helplessness forecasting models (default and bankruptcy), is the specificity of the model, which measures the model’s ability to predict unhealthy companies. The model based on Gradient Boosted Decision Tree has given poor performance on unbalanced data set. In addition, the sensitivity of the model also measures its accuracy among healthy companies. The Accuracy index also measures the overall accuracy of the model results, although both models have shown good performance, but the performance of the Random Forest model is clearly better. Finally, the Precision index measures the ability of the model to identify healthy companies in the data set. Due to the importance of data features in machine learning algorithms, models have been re-implemented in different cases by removing one or more features.

The ROC curve in Fig 2 is obtained using the features used in the Altman model (X1 to X5). In this case, the surface area under Fig 2 is equal to 0.88 and 0.83 for each of the models based on Random Forest and Gradient Boosted Decision Tree. As a result, it can be claimed that the research results of Barboza et al. [124] have been repeated in this research.

thumbnail
Download:
Fig 2. ROC curves for machine learning based on models with Altman features.

https://doi.org/10.1371/journal.pone.0292081.g002

In order to measure the importance of each feature, we remove one feature each time and implement the models. The ROC number after removing each feature individually is presented in Table 13.

thumbnail
Download:
Table 13. ROC number rfter removing each feature.

https://doi.org/10.1371/journal.pone.0292081.t013

As can be seen, based on the data in Table 13, the CPB feature, which is the change in the company’s P/B ratio, has a great impact on the explanatory power of machine learning algorithms. In another case, structural models have been implemented on all data consisting of company information from 2016 to 2021.

As can be seen in Fig 3, with the increase of the used data, the area under the ROC curve of Merton and Geske models improves and reaches 0.65 and 0.63. Considering the general increase in stock prices as a result of the historical growth of the Iranian capital market in the first half of 2020, and the reliance of structural models on stock prices and price fluctuations, these models have been used for the data from 2016 to 2019, and their performance results are shown in Fig 4.

thumbnail
Download:
Fig 3. ROC curves in implementation mode on data from 2016 to 2021.

https://doi.org/10.1371/journal.pone.0292081.g003

thumbnail
Download:
Fig 4. Implementation of structural models on data from 2016 to 2019.

https://doi.org/10.1371/journal.pone.0292081.g004

The obtained numbers as the area under Fig 4, similar to the previous case, were 0.65 and 0.63, which seems that the market developments in 2019 and 2020 have reduced the effectiveness of structural models.

Now we want to implement structural models on a different statistical sample. Due to not using data from 2016 to 2019 for structural models in the previous part, in this part, structural models for data from 2016 to 2021 are also used and the results are shown in Table 14.

thumbnail
Download:
Table 14. Prediction results of structural models for the entire data set.

https://doi.org/10.1371/journal.pone.0292081.t014

As it is clear from the results of the ROC curve, the overall performance of the structural models improves with the increase in the number of data. By increasing the sample size, the performance of Merton model becomes better than Geske model, which is confirmed by the Wilcoxon and T-tests.

5. Conclusions and future research directions

Default risk measurement and its role in credit risk measurement is an important role in measuring the more general concept of risk in companies in the capital market. On the other hand, the credit assessment of the loan applicant has always been the most important concern of the credit supply side, including banks, financial institutions, investment funds, suppliers of raw materials and other investors. Various models have been introduced to measure the default risk of companies, and in this research, four models, including the Merton model, the Geske model, and models based on random forest and Gradient Boosted Decision Tree, have been evaluated, and the results of each model and the parameters for measuring the predictive power of each mode is provided. In this research, we have checked the capacity of each model using the ROC curve. The total dataset used includes 306 companies (49 companies subject to Bankruptcy Article and 256 healthy companies). It should be noted that the information of 206 companies including 40 companies subject to Bankruptcy Article and 166 healthy companies were used as training data between 2016 and 2019, and also 100 companies including 9 companies subject to Bankruptcy Article and 91 healthy companies between the years 2020 to 2021 have been used as test data. Machine learning algorithms are implemented on all datasets using training data and test data. Then the structural models have been implemented on the data of 2020 and 2021 and finally the performance of the used models has been compared using the ROC curve. According to the results obtained from the surface area under the ROC figure, it can be concluded that the Merton model and the Geske model have performed better than random models with ROC scores of 0.54 to 0.65 and 0.58 to 0.63 respectively for different researches. Also, according to the conditions governing this research and according to the results obtained from the surface area under the ROC figure, it is evident that the Random Forest and Gradient Boosted Decision Tree models are very powerful models for predicting the default risk of companies, respectively, with a ROC score of 0.98 and 0.91 and accuracy score 0.96 and 0.87. Finally, for future studies, data envelopment analysis (DEA) method [133–169] and machine learning algorithms can be combined to performance prediction of listed companies in the Iranian capital market. Also, DEA as a powerful performance measurement tool can be applied for assessing credit risk management of companies [170–186].

Acknowledgments

The authors would like to thank the anonymous reviewers and the editor-in-chief for their constructive comments and suggestions.

References

  1. 1. Li X.; Wang J.; Yang CH. Risk prediction in financial management of listed companies based on optimized BP neural network under digital economy. Neural Computing and Applications 2023, Volume 35, Pages 2045–2058.
  2. 2. Peykani P.; Sargolzaei M.; Takaloo A.; Valizadeh S. The Effects of Monetary Policy on Macroeconomic Variables through Credit and Balance Sheet Channels: A Dynamic Stochastic General Equilibrium Approach. Sustainability 2023, 15(5), 4409;
  3. 3. Peykani P.; Sargolzaei M.; Botshekan M.H.; Oprean-Stan C.; Takaloo A. Optimization of Asset and Liability Management of Banks with Minimum Possible Changes. Mathematics 2023, 11(12), 2761;
  4. 4. Yan X.; Li Y.; Ming M.; Chong H. Impact of Nonstandard Default Risk of the Urban Investment and Development Companies on the Urban Investment Bond Market. Systems 2023, 11(2), 68;
  5. 5. Dorfleitner G.; Grebler J. Corporate social responsibility and systematic risk: international evidence. Journal of Risk Finance, Emerald Group Publishing Limited 2022, Vol. 23(1), Pages 85–120.
  6. 6. Duan Y.; Goodell J.W.; Li H.; Li X. Assessing machine learning for forecasting economic risk: Evidence from an expanded Chinese financial information set. Finance Research Letters 2022, Volume 46, Part A, 102273.
  7. 7. Landi G.C.; Iandolo F.; Renzi A.; Rey A. Embedding sustainability in risk management: The impact of environmental, social, and governance ratings on corporate financial risk. Corporate Social Responsibility and Environmental Management, John Wiley & Sons 2022, Vol. 29(4), Pages 1096–1107.
  8. 8. Lee CH.; Wang CH.; Ho SH. Financial aid and financial inclusion: Does risk uncertainty matter? Pacific-Basin Finance Journal 2022, Volume 71, 101700.
  9. 9. Elsayed A.; Gozgor G.; Lau CH. Risk transmissions between bitcoin and traditional financial assets during the COVID-19 era: The role of global uncertainties. International Review of Financial Analysis 2022, Volume 81, 102069.
  10. 10. Venturini A. Climate change, risk factors and stock returns: A review of the literature. International Review of Financial Analysis 2022, Volume 79, 101934.
  11. 11. Bai X.; Cheng L.; Iris C. Data-driven financial and operational risk management: Empirical evidence from the global tramp shipping industry. Transportation Research Part E: Logistics and Transportation Review 2022, Volume 158, 102617.
  12. 12. Ali M.; Alam N.; Khattak M.; Azmi W. Bank Risk-Taking and Legal Origin: What Do We Know about Dual Banking Economies? J. Risk Financial Manag. 2022, 15(5), 224;
  13. 13. Chen B.; Yang X.; Ma ZH. Fintech and Financial Risks of Systemically Important Commercial Banks in China: An Inverted U-Shaped Relationship. Sustainability 2022, 14(10), 5912;
  14. 14. ALrfai M.; Salleh D.; Waemustafa W. Empirical Examination of Credit Risk Determinant of Commercial Banks in Jordan. Risks 2022, 10(4), 85;
  15. 15. Ferrara L.; Mogliani M.; Sahuc J. High-frequency monitoring of growth at risk. International Journal of Forecasting 2022, Volume 38, Issue 2, Pages 582–595.
  16. 16. Dunbar K. Impact of the COVID-19 event on U.S. banks’ financial soundness. Research in International Business and Finance 2022, Volume 59, 101520.
  17. 17. Do T.D.; Pham H.A.; Thalassinos E.I.; Le H.A. The Impact of Digital Transformation on Performance: Evidence from Vietnamese Commercial Banks. J. Risk Financial Manag. 2022, 15(1), 21;
  18. 18. Cheng ZH.; Georgakakos K.; Spencer C.; Banks R. Numerical Modeling of Flash Flood Risk Mitigation and Operational Warning in Urban Areas. Water 2022, 14(16), 2494;
  19. 19. Liu H.; Huang W. Sustainable Financing and Financial Risk Management of Financial Institutions—Case Study on Chinese Banks. Sustainability 2022, 14(15), 9786;
  20. 20. Izcan D.; Bektas E. The Relationship between ESG Scores and Firm-Specific Risk of Eurozone Banks. Sustainability 2022, 14(14), 8619;
  21. 21. Rivera-Escobar O.; Escobar j.; Manotas D. Measurement of Systemic Risk in the Colombian Banking Sector. Risks 2022, 10(1), 22;
  22. 22. Abdul-Rahim R.; Bohari S.; Aman A.; Awang Z. Benefit–Risk Perceptions of FinTech Adoption for Sustainability from Bank Consumers’ Perspective: The Moderating Role of Fear of COVID-19. Sustainability 2022, 14(14), 8357;
  23. 23. Zou J.; Fu X.; Yang J.; Gong CH. Measuring Bank Systemic Risk in China: A Network Model Analysis. Systems 2022, 10(1), 14;
  24. 24. Leo M.; Sharma S.; Maddulety K. Machine Learning in Banking Risk Management: A Literature Review. Risks 2019, 7(1), 29;
  25. 25. Botshekan M.H.; Takaloo A.; Soureh R.H.; Abdollahi Poor M.S. Global Economic Policy Uncertainty (GEPU) and Non-Performing Loans (NPL) in Iran’s Banking System: Dynamic Correlation using the DCC-GARCH Approach. J. Mon. Ec. 2021, Vol. 16(2), Pages 187–212.
  26. 26. Katusiime L. COVID 19 and Bank Profitability in Low Income Countries: The Case of Uganda. J. Risk Financial Manag. 2021, 14(12), 588;
  27. 27. Umar M.; Ji X.; Mirza N.; Naqvi B. Carbon neutrality, bank lending, and credit risk: Evidence from the Eurozone. Journal of Environmental Management 2021, Volume 296, 113156. pmid:34225048
  28. 28. Rastogi S.; Gupte R.; Meenakshi R. A Holistic Perspective on Bank Performance Using Regulation, Profitability, and Risk-Taking with a View on Ownership Concentration. J. Risk Financial Manag. 2021, 14(3), 111;
  29. 29. Takaloo A.; Abdollahipour M.S.; Sargolzaei M. Shariah Compliance Framework in Compliance with Corporate Governance Principles in Malaysia’s Banking System. Quarterly Studies in Banking Management and Islamic Banking 2022, Volume 8, Issue 18, Pages 29–62.
  30. 30. Olmo B.; Saiz M.; Azofra S. Sustainable Banking, Market Power, and Efficiency: Effects on Banks’ Profitability and Risk. Sustainability 2021, 13(3), 1298;
  31. 31. Sargolzaei M.; Takaloo A.; Seyedian S.M. The effect of stock market liquidity on the market power of banks. The Journal of Economic Policy 2023, Volume 14, Issue 28, Pages 315–344.
  32. 32. Misman F.N.; Bhatti M.I. The Determinants of Credit Risk: An Evidence from ASEAN and GCC Islamic Banks. J. Risk Financial Manag. 2020, 13(5), 89;
  33. 33. Nguyen K.N. Revenue Diversification, Risk and Bank Performance of Vietnamese Commercial Banks. J. Risk Financial Manag. 2019, 12(3), 138;
  34. 34. Al Rahahleh N.; Bhatti M.I.; Misman F.N. Developments in Risk Management in Islamic Finance: A Review. J. Risk Financial Manag. 2019, 12(1), 37;
  35. 35. Meles A.; Salerno D.; Sampagnaro G.; Verdoliva V.; Zhang J. The influence of green innovation on default risk: Evidence from Europe. International Review of Economics and Finance 2023, Volume 84, Pages 692–710.
  36. 36. Bonaccolto G.; Borri N.; Consiglio A. Breakup and default risks in the great lockdown. Journal of Banking and Finance 2023, Volume 147, 106308.
  37. 37. Spatareanu M.; Manole V.; Kabiri A.; Roland I. Bank default risk propagation along supply chains: Evidence from the U.K. International Review of Economics and Finance 2023, Volume 84, Pages 813–831.
  38. 38. Yu W.; Zhang Y.; Du K.; Wu Y. Does the Quality of Director Fusion Raise the Risk of Corporate Debt Default? Sustainability 2023, 15(2), 1698;
  39. 39. Mirza n.; Afzal A.; Umar M.; Skare M. The impact of green lending on banking performance: Evidence from SME credit portfolios in the BRIC. Economic Analysis and Policy 2023, Volume 77, Pages 843–850.
  40. 40. Jermann U.; Xiang H. Dynamic banking with non-maturing deposits. Journal of Economic Theory 2023, Volume 209, 105644.
  41. 41. Soenen N.; Vennet R.V. Determinants of European banks’ default risk. Finance Research Letters 2022, Volume 47, Part A, 102557.
  42. 42. Azeem Khan A. Ahmad W. Fresh evidence on the relationship between market power and default risk of Indian banks. Finance Research Letters 2022, Volume 46, Part A, 102360.
  43. 43. García J.; Herrero B.; Morillas F. Corporate board and default risk of financial firms. Economic Research 2022, Volume 35, Issue 1.
  44. 44. Karaye A.I.; Ahmad-Zaluki N.A.; Badru B.O. The Effect of Credit Committee Characteristics on Bank Asset Quality in Nigeria. Financial Markets, Institutions and Risks 2022, Volume 6, Issue 2, Pages 60–74.
  45. 45. Dibooglu S.; Cevik E.; Al Tamimi H. Credit default risk in Islamic and conventional banks: Evidence from a GARCH option pricing model. Economic Analysis and Policy 2022, Volume 75, Pages 396–411.
  46. 46. Soenen N.; Vennet R.V. ECB monetary policy and bank default risk. Journal of International Money and Finance 2022, Volume 122, 102571.
  47. 47. Xuezhou W.; Hussain R.Y.; Salameh A.A.; Hussain H.; Burhan Khan A.; Fareed M. Does Firm Growth Impede or Expedite Insolvency Risk? A Mediated Moderation Model of Leverage Maturity and Potential Fixed Collaterals. Frontiers in Environmental Science 2022, Volume 10,
  48. 48. Nguyen D.; Nguyen C.; Dang L. Uncertainty and corporate default risk: Novel evidence from emerging markets. Journal of International Financial Markets, Institutions and Money 2022, Volume 78, 101571.
  49. 49. Gomez-Gonzalez J.E.; Valencia O.M.; Sánchez G.A. How fiscal rules can reduce sovereign debt default risk. Emerging Markets Review 2022, Volume 50, 100839.
  50. 50. Zhang X.; Zhao Y.; Yao X. Forecasting corporate default risk in China. International Journal of Forecasting 2022, Volume 38, Issue 3, Pages 1054–1070.
  51. 51. Burhan Khan A.; Fareed M.; Salameh A.A.; Hussain H. Financial Innovation, Sustainable Economic Growth, and Credit Risk: A Case of the ASEAN Banking Sector. Frontiers in Environmental Science 2021, 9(1):1–10.
  52. 52. Joseph A.D. Emerging Market Default Risk Charge Model. J. Risk Financial Manag. 2023, 16(3), 194;
  53. 53. Wang W.; Li Q.; Li Q.; Xu S. Robust Optimal Investment Strategies with Exchange Rate Risk and Default Risk. Mathematics 2023, 11(6), 1550;
  54. 54. Suganda T.R.; Kim J. An Empirical Study on the Relationship between Corporate Social Responsibility and Default Risk: Evidence in Korea. Sustainability 2023, 15(4), 3644;
  55. 55. Yang Z. GameStop or Game Just Started? Leveling the Playing Field for Social Media Meme Investors to Rebuild the Public’s Trust. J. Risk Financial Manag. 2023, 16(1), 13;
  56. 56. Saci K.; Mansour W. Risk Sharing, SMEs’ Financial Strategy, and Lending Guarantee Technology. Risks 2023, 11(2), 33;
  57. 57. Wang Y.; Shen J.; Pan J.; Chen T. A Credit Risk Contagion Intensity Model of Supply Chain Enterprises under Different Credit Modes. Sustainability 2022, 14(20), 13518;
  58. 58. Li X.; Tian X. Research on SMEs’ Reputation Mechanism and Default Risk Based on Investors’ Financial Participation. Sustainability 2022, 14(21), 14329;
  59. 59. Gao Q. Systemic Risk Analysis of Multi-Layer Financial Network System Based on Multiple Interconnections between Banks, Firms, and Assets. Entropy 2022, 24(9), 1252; pmid:36141138
  60. 60. Juhász P.; Felföldi-Szűcs N. Financing Cooperative Supply Chain Members—The Bank’s Perspective. Risks 2022, 10(7), 139;
  61. 61. Li X.; Lu T.; Lin J. Bank Interest Margin and Green Lending Policy under Sunflower Management. Sustainability 2022, 14(14), 8643;
  62. 62. Rikhotso P.M.; Simo-Kengne B.D. Dependence Structures between Sovereign Credit Default Swaps and Global Risk Factors in BRICS Countries. J. Risk Financial Manag. 2022, 15(3), 109;
  63. 63. Wang CH.; Xiao ZH. A Deep Learning Approach for Credit Scoring Using Feature Embedded Transformer. Appl. Sci. 2022, 12(21), 10995;
  64. 64. Sidhu A.; Rastogi S.; Gupte R.; Bhimavarapu V. Impact of Liquidity Coverage Ratio on Performance of Select Indian Banks. J. Risk Financial Manag. 2022, 15(5), 226;
  65. 65. Kristóf T. Sovereign Default Forecasting in the Era of the COVID-19 Crisis. J. Risk Financial Manag. 2021, 14(10), 494;
  66. 66. Zheng H.; Zhang ZH. Analyzing Characteristics and Implications of the Mortgage Default of Agricultural Land Management Rights in Recent China Based on 724 Court Decisions. Land 2021, 10(7), 729;
  67. 67. Weng ZH.; Luo P. Supervision of the Default Risk of Online Car-Hailing Platform from an Evolutionary Game Perspective. Sustainability 2021, 13(2), 555;
  68. 68. Choudhury T.; Scagnelli S.; Yong J.; Zhang ZH. Non-Traditional Systemic Risk Contagion within the Chinese Banking Industry. Sustainability 2021, 13(14), 7954;
  69. 69. Ji X.; Yu L.; Fu J. Evaluating Personal Default Risk in P2P Lending Platform: Based on Dual Hesitant Pythagorean Fuzzy TODIM Approach. Mathematics 2020, 8(1), 8;
  70. 70. Kim H.; Cho H.; Ryu D. Corporate Default Predictions Using Machine Learning: Literature Review. Sustainability 2020, 12(16), 6325;
  71. 71. Rahman H.; Yousaf M.; Tabassum M. Bank-Specific and Macroeconomic Determinants of Profitability: A Revisit of Pakistani Banking Sector under Dynamic Panel Data Approach. Int. J. Financial Stud. 2020, 8(3), 42;
  72. 72. Habermann F.; Fischer F. Corporate Social Performance and the Likelihood of Bankruptcy: Evidence from a Period of Economic Upswing. Journal of Business Ethics 2023, Volume 182, Pages 243–259.
  73. 73. Zhao Y.; Lin D. Prediction of Micro- and Small-Sized Enterprise Default Risk Based on a Logistic Model: Evidence from a Bank of China. Sustainability 2023, 15(5), 4097;
  74. 74. Li Y.; Ou J.; Gu CH. Buyer guarantee and bailout in supplier finance with bankruptcy cost. European Journal of Operational Research 2023, Volume 305, Issue 1, Pages 287–299.
  75. 75. Aliano M.; Cananà L.; Cestari G.; Ragni S. A Dynamical Model with Time Delay for Risk Contagion. Mathematics 2023, 11(2), 425;
  76. 76. Ji Y.; Shi L.; Zhang SH. Digital finance and corporate bankruptcy risk: Evidence from China. Pacific-Basin Finance Journal 2022, Volume 72, 101731.
  77. 77. Senarath S.; Rajapakse P.; Robbé J.; Wickremeratne N.; Subasinghage M. Being Naked—et Quo hinc?: Developing a ‘Skin-in-the-Game’ Solution for Credit Default Swaps. Int. J. Financial Stud. 2022, 10(4), 94;
  78. 78. Matenda F.; Sibanda M. Determinants of Default Probability for Audited and Unaudited SMEs under Stressed Conditions in Zimbabwe. Economies 2022, 10(11), 274;
  79. 79. Tran K.; Le H.; Nguyen T.; Nguyen D. Explainable Machine Learning for Financial Distress Prediction: Evidence from Vietnam. Data 2022, 7(11), 160;
  80. 80. Costa M.; Lisboa I.; Gameiro A. Is the Financial Report Quality Important in the Default Prediction? SME Portuguese Construction Sector Evidence. Risks 2022, 10(5), 98;
  81. 81. Trinh H.; Nguyen C.; Hao W.; Wongchoti U. Does stock liquidity affect bankruptcy risk? DID analysis from Vietnam. Pacific-Basin Finance Journal 2021, Volume 69, 101634.
  82. 82. Mari C.; Marra M. Valuing firm’s financial flexibility under default risk and bankruptcy costs: a WACC based approach. International Journal of Managerial Finance, Emerald Group Publishing Limited 2019, Vol. 15(5), Pages 688–699.
  83. 83. Zhao S.; Lu X. Guarantee Mechanism in Accounts Receivable Financing with Demand Uncertainty. Sustainability 2023, 15(3), 2192;
  84. 84. Xu R.; Guo T.; Zhao H. Research on the Path of Policy Financing Guarantee to Promote SMEs’ Green Technology Innovation. Mathematics 2022, 10(4), 642;
  85. 85. Yoshino N.; Taghizadeh-Hesary F. Optimal credit guarantee ratio for small and medium-sized enterprises’ financing: Evidence from Asia. Economic Analysis and Policy 2019, Volume 62, Pages 342–356.
  86. 86. Wilcox J.A.; Yasuda Y. Government guarantees of loans to small businesses: Effects on banks’ risk-taking and non-guaranteed lending. Journal of Financial Intermediation 2019, Volume 37, Pages 45–57.
  87. 87. Desogus M.; Venturi B. Stability and Bifurcations in Banks and Small Enterprises—A Three-Dimensional Continuous-Time Dynamical System. J. Risk Financial Manag. 2023, 16(3), 171;
  88. 88. Zhang Q.; Yang D.; Qin J. Multi-Party Evolutionary Game Analysis of Accounts Receivable Financing under the Application of Central Bank Digital Currency. J. Theor. Appl. Electron. Commer. Res. 2023, 18(1), 394–415;
  89. 89. Chen CH.; Huang H.; Zhao B.; Shu D.; Wang Y. The Research of AHP-Based Credit Rating System on a Blockchain Application. Electronics 2023, 12(4), 887;
  90. 90. Sun K. Do Rating Change Announcements Transfer Effective Information? Test on the Effectiveness and Sustainability of Credit Rating in China. Sustainability 2022, 14(21), 14086;
  91. 91. Zuo M.; Wu T. Does Environmental Credit Rating Promote Green Innovation in Enterprises? Evidence from Heavy Polluting Listed Companies in China. Int. J. Environ. Res. Public Health 2022, 19(20), 13617; pmid:36294195
  92. 92. Jelinek S.; Milošević P.; Rakićević A.; Poledica A.; Petrović B. A Novel IBA-DE Hybrid Approach for Modeling Sovereign Credit Ratings. Mathematics 2022, 10(15), 2679;
  93. 93. Takawira O.; Mwamba J. Sovereign Credit Ratings Analysis Using the Logistic Regression Model. Risks 2022, 10(4), 70;
  94. 94. Estran R.; Souchaud A.; Abitbol D. Using a genetic algorithm to optimize an expert credit rating model. Expert Systems with Applications 2022, Volume 203, 117506.
  95. 95. Jiang W.; Xu L.; Chen ZH.; Govindan K.; Chin K. Financing equilibrium in a capital constrained supply Chain: The impact of credit rating. Transportation Research Part E: Logistics and Transportation Review 2022, Volume 157, 102559.
  96. 96. Ubarhande P.; Chandani A.; McMillan D. Elements of Credit Rating: A Hybrid Review and Future Research Agenda. Cogent Business and Management 2021, Volume 8, Issue 1.
  97. 97. Mansoor M.; Ellahi N.; Hassan A.; Malik Q.; Waheed A.; Ullah N. Corporate Governance, Shariah Governance, and Credit Rating: A Cross-Country Analysis from Asian Islamic Banks. J. Open Innov. Technol. Mark. Complex. 2020, 6(4), 170;
  98. 98. Baofeng SH.; Bi W.; Xue ZH.; Yizhe D. Credit rating and microfinance lending decisions based on loss given default (LGD). Finance Research Letters 2019, Volume 30, Pages 124–129.
  99. 99. Song S.; Tang D.; Xu G.; Yin X. An analytical GARCH valuation model for spread options with default risk. International Review of Economics and Finance 2023, Volume 83, Pages 1–20.
  100. 100. Viganò F. The Climate Financialization Trap: Claiming for Public Action. Sustainability 2023, 15(6), 4841;
  101. 101. Chatoro M.; Mitra S.; Pantelous A.; Shao J.Catastrophe bond pricing in the primary market: The issuer effect and pricing factors. International Review of Financial Analysis 2023, Volume 85, 102431.
  102. 102. Hu CH.; Liu M.; Jiang W. The Effect of CDS Trading on Product Market Competition: Evidence from 10-K Filings. J. Risk Financial Manag. 2023, 16(3), 207;
  103. 103. Wu F.; Ding D.; Yin J.; Lu W.; Yuan G. Total Value Adjustment of Multi-Asset Derivatives under Multivariate CGMY Processes. Fractal Fract. 2023, 7(4), 308;
  104. 104. Naifar N.; Aljarba SH. Does Geopolitical Risk Matter for Sovereign Credit Risk? Fresh Evidence from Nonlinear Analysis. J. Risk Financial Manag. 2023, 16(3), 148;
  105. 105. Jones E.; Mason S.; Rosenfeld E. Contingent Claims Analysis of Corporate Capital Structures: An Empirical Investigation. Journal of Finance 1984, Vol. 39, Issue 3, Pages 611–625.
  106. 106. Ogden J. Determinants Of The Ratings And Yields On Corporate Bonds: Tests Of The Contingent Claims Model. Journal of Financial Research, Southern Finance Association;Southwestern Finance Association 1987, Vol. 10(4), Pages 329–340,
  107. 107. Zhou CH. A jump-diffusion approach to modeling credit risk and valuing defaultable securities. Finance and Economics Discussion Series 1997, Pages 1–45.
  108. 108. Lyden S.; Saraniti D. An Examination of the Classical Theory of Corporate Security Valuation. SSRN Electronic Journal 2001,
  109. 109. Ho Eom Y.; Helwege J.; Huang J. Structural Models of Corporate Bond Pricing: An Empirical Analysis. The Review of Financial Studies 2004, Vol. 17, No. 2, Pages 499–544.
  110. 110. Jarrow R.; Van Deventer D.; Wang X. A robust test of Merton’s structural model for credit risk. Journal of Risk 2003, Vol. 6(1), Pages 39–58,
  111. 111. Schäfer R.; Sjölin M.; Sundin A.; Wolanski M.; Guhr T. Credit risk—A structural model with jumps and correlations. Physica A: Statistical Mechanics and its Applications 2007, Volume 383, Issue 2, Pages 533–569.
  112. 112. Schaefer S.; Strebulaev I. Structural models of credit risk are useful: Evidence from hedge ratios on corporate bonds. Journal of Financial Economics 2008, Volume 90, Issue 1, Pages 1–19.
  113. 113. Khansari R.; Fallahshams M. Appraising the Use of KMV Model in Predicting Default of Companies Listed in Tehran Stock Exchange. Financial Research Journal 2010, Volume 11, Issue 28, Number 28.
  114. 114. Arora N.; Bohn J.; Zhu F. Reduced-Form versus Structural Models of Credit Risk: A Case Study of Three Models: The Credit Market Handbook: Advanced Modeling Issues. Wiley Online Library 2012,
  115. 115. Liang G.; Jiang L. A modified structural model for credit risk. IMA Journal of Management Mathematics 2012, Vol. 23(2)
  116. 116. Gadzo S.; Kportorgbi H.; Gatsi J.; Murray L. Credit risk and operational risk on financial performance of universal banks in Ghana: A partial least squared structural equation model (PLS SEM) approach. Cogent Economics & Finance 2019, Volume 7, Issue 1.
  117. 117. Pasricha P.; Lu X.; Zhu S. A note on the calculation of default probabilities in “Structural credit risk modeling with Hawkes jump–diffusion processes”. Journal of Computational and Applied Mathematics 2021, Volume 381, 1113037.
  118. 118. Khandani A.; Kim A.; Lo A. Consumer credit-risk models via machine-learning algorithms. Journal of Banking and Finance 2010, Vol. 34(11), Pages 2767–2787.
  119. 119. Brown I.; Mues CH. An experimental comparison of classification algorithms for imbalanced credit scoring data sets. Expert Systems with Applications 2012, Vol. 39(3), Pages 3446–3453.
  120. 120. Fitzpatrick T.; Mues CH. An empirical comparison of classification algorithms for mortgage default prediction: evidence from a distressed mortgage market. European Journal of Operational Research, Elsevier 2014, Vol. 249(2), Pages 427–439.
  121. 121. Panahi H.; Asadzadeh A.; Jalili Marand A.Use of Combined Approach of Support Vector Machine and Feature Selection for Financial Distress Prediction of Listed Companies in Tehran Stock Exchange Market. Financial Research Journal 2014, Volume 16, Issue 1, Number 1, Pages 57–76.
  122. 122. Chakraborty, CH.; Joseph A. Machine learning at central banks. Bank of England working papers 674, Bank of England 2017.
    • 123. Jones S. Corporate bankruptcy prediction: a high dimensional analysis. Review of Accounting Studies 2017, Vol. 22(4), Pages 1366–1422.
    • 124. Barboza F.; Kimura H.; Altman E. Machine learning models and bankruptcy prediction. Expert Systems With Applications 2017, Volume 83, Pages 405–417.
    • 125. Fallahpour S.; Norouzian Lakvan E.; Hendijani Zadeh M. Use of Combined Approach of Support Vector Machine and Feature Selection for Financial Distress Prediction of Listed Companies in Tehran Stock Exchange Market. Financial Research Journal 2017, Volume 19, Issue 1, Pages 139–156.
    • 126. Fuster A.; Goldsmith-Pinkham P.; Ramadorai T.; Walther A. Predictably Unequal? The Effects of Machine Learning on Credit Markets. Journal of Finance 2018, Vol. 77(1), Pages 5–47.
    • 127. Zhu L.; Qiu D.; Ergu D.; Ying C.; Liu K. A study on predicting loan default based on the random forest algorithm. Procedia Computer Science 2019, Vol. 162, Pages 503–513.
    • 128. Son H.; Hyun C.; Phan D.; Hwang H. Data analytic approach for bankruptcy prediction. Expert Systems with Applications 2019, Volume 138, 112816.
    • 129. Deng SH.; Wang CH.; Wang M.; Sun ZH. A gradient boosting decision tree approach for insider trading identification: An empirical model evaluation of China stock market. Applied Soft Computing 2019, Volume 83, 105652.
    • 130. Sigrist F.; Hirnschall CH. Grabit: Gradient tree-boosted Tobit models for default prediction. Journal of Banking & Finance 2019, Vol. 102, Pages 177–192.
    • 131. Acharya S.; Pustokhina I.; Pustokhin D.; Geetha B. An improved gradient boosting tree algorithm for financial risk management. Knowledge Management Research & Practice 2021,
    • 132. Yıldırım M.; Okay F.; Özdemir S. Big data analytics for default prediction using graph theory. Expert Systems with Applications 2021, Volume 176, 114840.
    • 133. Abad C., Thore S. A., & Laffarga J. (2004). Fundamental analysis of stocks by two‐stage DEA. Managerial and Decision Economics, 25(5), 231–241.
    • 134. Peykani P.; Mohammadi E.; Pishvaee M.S.; Rostamy-Malkhalifeh M.; Jabbarzadeh A. A novel fuzzy data envelopment analysis based on robust possibilistic programming: Possibility, necessity and credibility-based approaches. RAIRO-Oper. Res. 2018, 52, 1445–1463.
    • 135. Arabjazi N., Rostamy-Malkhalifeh M., Hosseinzadeh Lotfi F., & Behzadi M. H. Stochastic sensitivity analysis in data envelopment analysis. Fuzzy Optim. Modeling J. 2021, 2, 52–64.
    • 136. Peykani P.; Mohammadi E.; Emrouznejad A.; Pishvaee M.S.; Rostamy-Malkhalifeh M. Fuzzy data envelopment analysis: An adjustable approach. Expert Syst. Appl. 2019, 136, 439–452.
    • 137. Arabjazi N., Rostamy-Malkhalifeh M., Hosseinzadeh Lotfi F., & Behzadi M. H. Determining the exact stability region and radius through efficient hyperplanes. Iran. J. Manag. Stud. 2022, 15, 287–303.
    • 138. Arabjazi N., Rostamy-Malkhalifeh M., Hosseinzadeh Lotfi F., & Behzadi M. H. Stability analysis with general fuzzy measure: An application to social security organizations. PLoS ONE 2022, 17, e0275594. pmid:36264849
    • 139. Peykani P., & Gheidar-Kheljani J. Performance appraisal of research and development projects value-chain for complex products and systems: the fuzzy three-stage DEA approach. Journal of New Researches in Mathematics, 2020, 6, 41–58.
    • 140. Peykani P.; Mohammadi E.; Farzipoor Saen R.; Sadjadi S.J.; Rostamy-Malkhalifeh M. Data envelopment analysis and robust optimization: A review. Expert Syst. 2020, 37, e12534.
    • 141. Bortoluzzi M., Furlan M., & dos Reis Neto J. F. (2022). Assessing the impact of hydropower projects in Brazil through data envelopment analysis and machine learning. Renewable Energy, 200, 1316–1326.
    • 142. Duras T., Javed F., Månsson K., Sjölander P., & Söderberg M. (2023). Using machine learning to select variables in data envelopment analysis: Simulations and application using electricity distribution data. Energy Economics, 120, 106621.
    • 143. Guerrero N. M., Aparicio J., & Valero-Carreras D. (2022). Combining Data Envelopment Analysis and Machine Learning. Mathematics, 10(6), 909.
    • 144. Peykani P.; Mohammadi E.; Jabbarzadeh A.; Rostamy-Malkhalifeh M.; Pishvaee M.S. A novel two-phase robust portfolio selection and optimization approach under uncertainty: A case study of Tehran stock exchange. PLoS ONE 2020, 15, e0239810. pmid:33045010
    • 145. Peykani P.; Farzipoor Saen R.; Seyed Esmaeili F.S.; Gheidar-Kheljani J. Window data envelopment analysis approach: A review and bibliometric analysis. Expert Syst. 2021, 38, e12721.
    • 146. Jothimani D., Shankar R., & Yadav S. S. (2017). A PCA-DEA framework for stock selection in Indian stock market. Journal of Modelling in Management, 12(3), 386–403.
    • 147. Kehinde T. O., Chan F. T., & Chung S. H. (2023). Scientometric review and analysis of recent approaches to stock market forecasting: Two decades survey. Expert Systems with Applications, 213, 119299.
    • 148. Li J., Gao H., Li Y., Jin X., & Liang L. (2022). Stock efficiency evaluation based on multiple risk measures: a DEA-like envelopment approach. Journal of Systems Science and Complexity, 35(4), 1480–1499.
    • 149. Lim S., Oh K. W., & Zhu J. (2014). Use of DEA cross-efficiency evaluation in portfolio selection: An application to Korean stock market. European Journal of Operational Research, 236(1), 361–368.
    • 150. Mohtashami A., & Ghiasvand B. M. (2020). Z-ERM DEA integrated approach for evaluation of banks & financial institutes in stock exchange. Expert Systems with Applications, 147, 113218.
    • 151. Peykani P.; Mohammadi E.; Emrouznejad A. An adjustable fuzzy chance-constrained network DEA approach with application to ranking investment firms. Expert Syst. Appl. 2021, 166, 113938.
    • 152. Peykani P.; Namakshenas M.; Arabjazi N.; Shirazi F.; Kavand N. Optimistic and pessimistic fuzzy data envelopment analysis: Empirical evidence from Tehran stock market. Fuzzy Optim. Modeling J. 2021, 2, 12–21.
    • 153. Peykani P.; Seyed Esmaeili F.S. Malmquist productivity index under fuzzy environment. Fuzzy Optim. Modeling J. 2021, 2, 10–19.
    • 154. Nouri M., Mohammadi E., & Rahmanipour M. A novel efficiency ranking approach based on goal programming and data envelopment analysis for the evaluation of Iranian banks. Int. J. Data Envel. Anal. 2019, 7, 57–80.
    • 155. Pan W., Qian W., Xia Y., Liu Z., & Liu Y. (2022). Analysis of Economic Benefits of China’s Joint-Stock Commercial Banks based on Two-Stage Network DEA and DEA-Malmquist. Forest Chemicals Review, 448–459.
    • 156. Rasoulzadeh M., Edalatpanah S. A., Fallah M., & Najafi S. E. (2022). A multi-objective approach based on Markowitz and DEA cross-efficiency models for the intuitionistic fuzzy portfolio selection problem. Decision Making: Applications in Management and Engineering, 5(2), 241–259.
    • 157. Peykani P.; Hosseinzadeh Lotfi F.; Sadjadi S.J.; Ebrahimnejad A.; Mohammadi E. Fuzzy chance-constrained data envelopment analysis: A structured literature review, current trends, and future directions. Fuzzy Optim. Decis. Mak. 2022, 21, 197–261.
    • 158. Peykani P.; Namazi M.; Mohammadi E. Bridging the knowledge gap between technology and business: An innovation strategy perspective. PLoS ONE 2022, 17, e0266843. pmid:35421135
    • 159. Seyed Esmaeili F.S. The efficiency of MSBM model with imprecise data (interval). Int. J. Data Envel. Anal. 2014, 2, 343–350.
    • 160. Seyed Esmaeili F.S.; Rostamy-Malkhalifeh M.; Hosseinzadeh Lotfi F. Two-stage network DEA model under interval data. Math. Anal. Convex Optim. 2020, 1, 103–108.
    • 161. Peykani P.; Emrouznejad A.; Mohammadi E.; Gheidar‐Kheljani J. A novel robust network data envelopment analysis approach for performance assessment of mutual funds under uncertainty. Ann. Oper. Res. 2022, 1–27.
    • 162. Peykani P., Gheidar-Kheljani J., Farzipoor Saen R., & Mohammadi E. Generalized robust window data envelopment analysis approach for dynamic performance measurement under uncertain panel data. Oper Res Int J. 2022, 22, 5529–5567.
    • 163. Seyed Esmaeili F.S.; Rostamy-Malkhalifeh M.; Hosseinzadeh Lotfi F. A hybrid approach using data envelopment analysis, interval programming and robust optimisation for performance assessment of hotels under uncertainty. Int. J. Manag. Decis. Mak. 2021, 20, 308–322.
    • 164. Seyed Esmaeili F.S.; Rostamy-Malkhalifeh M.; Hosseinzadeh Lotfi F. Interval network Malmquist productivity index for examining productivity changes of insurance companies under data uncertainty: A case study. J. Math. Ext. 2022, 16, 9.
    • 165. Peykani P.; Memar-Masjed E.; Arabjazi N.; Mirmozaffari M. Dynamic performance assessment of hospitals by applying credibility-based fuzzy window data envelopment analysis. Healthcare 2022, 10, 876. pmid:35628013
    • 166. Peykani P.; Seyed Esmaeili F.S.; Mirmozaffari M.; Jabbarzadeh A.; Khamechian M. Input/Output Variables Selection in Data Envelopment Analysis: A Shannon Entropy Approach. Mach. Learn. Knowl. Extr. 2022, 4, 688–699.
    • 167. Yan Z., Zhou W., Wang Y., & Chen X. (2022). Comprehensive Analysis of Grain Production Based on Three-Stage Super-SBM DEA and Machine Learning in Hexi Corridor, China. Sustainability, 14(14), 8881.
    • 168. Zhang Z., Xiao Y., & Niu H. (2022). DEA and Machine Learning for Performance Prediction. Mathematics, 10(10), 1776.
    • 169. Zhang Z., Xiao Y., Fu Z., Zhong K., & Niu H. (2022). A study on early warnings of financial crisis of Chinese listed companies based on DEA–SVM model. Mathematics, 10(12), 2142.
    • 170. Boussemart J. P., Leleu H., Shen Z., Vardanyan M., & Zhu N. (2019). Decomposing banking performance into economic and credit risk efficiencies. European Journal of Operational Research, 277(2), 719–726.
    • 171. Bruni M. E., Beraldi P., & Iazzolino G. (2014). Lending decisions under uncertainty: a DEA approach. International Journal of Production Research, 52(3), 766–775.
    • 172. Dahooie J. H., Hajiagha S. H. R., Farazmehr S., Zavadskas E. K., & Antucheviciene J. (2021). A novel dynamic credit risk evaluation method using data envelopment analysis with common weights and combination of multi-attribute decision-making methods. Computers & Operations Research, 129, 105223.
    • 173. Feruś A. (2008). The DEA method in managing the credit risk of companies. Ekonomika, 84, 109–118.
    • 174. Feruś A. (2010). The Application of DEA Method in Evaluating Credit Risk of Companies. Contemporary Economics, 4(4), 107–114.
    • 175. Feruś A. (2014). The application of data envelopment analysis method in managing companies’ credit risk. Business and Economic Horizons, 10(1), 60–69.
    • 176. Ghafoorian H., Norhan N., Abubakar M. N., & Nodeh F. M. (2013). Efficiency Considering Credit Risk in Banking Industry, using two-stage DEA. Journal of Social and Development Sciences, 4(8), 356–360.
    • 177. Iazzolino G., Bruni M. E., & Beraldi P. (2013). Using DEA and financial ratings for credit risk evaluation: an empirical analysis. Applied Economics Letters, 20(14), 1310–1317.
    • 178. Keramati M. A., & Shaeri M. (2014). Assessment of credit risk management and managerial efficiency of banks using data envelopment analysis (DEA) network. Biological Forum, 6(2), 320–328.
    • 179. Li R., Li L., & Zou P. (2020). Credit risk shocks and banking efficiency: a study based on a bootstrap-DEA model with nonperforming loans as bad output. Journal of Economic Studies, 48(1), 1–19.
    • 180. Lu S. L., Lee K. J., & Zou M. L. (2012). How to gauge credit risk: an investigation based on data envelopment analysis and the Markov chain model. Applied Financial Economics, 22(11), 887–897.
    • 181. Paradi J. C., Asmild M., & Simak P. C. (2004). Using DEA and worst practice DEA in credit risk evaluation. Journal of Productivity Analysis, 21, 153–165.
    • 182. Pasiouras F. (2008). Estimating the technical and scale efficiency of Greek commercial banks: the impact of credit risk, off-balance sheet activities, and international operations. Research in International Business and Finance, 22(3), 301–318.
    • 183. Pastor J. M. (2002). Credit risk and efficiency in the European banking system: A three-stage analysis. Applied Financial Economics, 12(12), 895–911.
    • 184. Psillaki M., Tsolas I. E., & Margaritis D. (2010). Evaluation of credit risk based on firm performance. European Journal of Operational Research, 201(3), 873–881.
    • 185. Tsolas I. E. (2015). Firm credit risk evaluation: a series two-stage DEA modeling framework. Annals of Operations Research, 233, 483–500.
    • 186. Tsolas I. E., & Charles V. (2015). Incorporating risk into bank efficiency: A satisficing DEA approach to assess the Greek banking crisis. Expert Systems with Applications, 42(7), 3491–3500.