|
Identification and control of credit risk in banks relying on new monitoring technologies with K-MEANS clustering algorithm and support vector machine
|
Maryam Mashrooti1   , Ali Mohammadi2 , Mehdi Mohammadi3 |
1- PhD student of accounting department, Zanjan branch, Islamic Azad University, Zanjan, Iran.
2- Assistant Professor, Department of Accounting, Zanjan Branch, Islamic Azad University, Zanjan, Iran. , Ali.mohammadi@iauz.ac.ir
3- Assistant Professor, Department of Accounting, Zanjan Branch, Islamic Azad University, Zanjan, Iran. |
|
|
Abstract: (19 Views) |
Today, the only cause of bank failure is credit risk. If the customer does not repay his obligations on time. These facilities are in the form of outstanding bank claims and this causes disruption in the banking system and as a result in the country's economy. New surveillance technologies are still in their infancy, but they are gaining recognition and attention. The latest generations of big data or artificial intelligence that regulatory agencies are using have emerged recently. Therefore, the main issue of this research is the control of credit risk in financial institutions with emphasis on the new supervisory technologies of Saptech and Regtech with new approaches of data mining.
Considering the quantitative nature of the research and the use of data mining to validate bank customers, this research is data-oriented. The main basis of the present research is the discovery of knowledge from bank databases. In this research, after collecting the data of the former customers of the bank from the relevant database and after that, refining the data, it is done to identify the influencing variables in the rating of the customers, which is done through the review of previous scientific researches. , is done. In the next step, by using K-means clustering techniques and support vector machine and with the help of relevant software, customers are classified based on their characteristics and their behavior is predicted. For the purpose of credit rating, the analysis of the information related to the real customers of Tejarat Bank and Saman Bank of Tehran, who are the trustees of implementing Saptek and Regtek, is used, and with the Cochran formula, 230 samples of customers with accounts leading to the year 1400 to 1401 are used. were chosen.
Among the research variables, "collateral value, interest rate and inflation rate" had the greatest impact on credit risk. The results showed that the accuracy of the selected techniques models in this research was very good and these models were able to recognize 81.02% of risky and non-risky customers on average. Also, according to the results, feature selection in all techniques has increased the prediction accuracy. The support vector machine (SVM) technique used in this research has the highest accuracy in all models, and by selecting the features, the accuracy of this model has increased compared to the base model, and it has the highest accuracy (81.58%) in Among all the techniques. |
Article number: 16 |
| Keywords: credit risk, risk management, dishonour a bill, supervisory technology, data mining |
|
|
Full-Text [PDF 1718 kb]
(8 Downloads)
|
Article type: Research |
Subject:
General
Received: 2024/10/31 | Accepted: 2025/01/19 | Published: 2026/01/30
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
