Credit Risk assessment Model For Maskan Bank Legal Customers Using Logistic Regression

  • Sobhan Hoosein Beigy et al.


Providing credit facilities to clients can be regarded as one of the most important tasks of banks. Banks in
each country, after collecting financial resources, allocate these resources to different economic sectors.
In fact, this action by the banks will strengthen the various economic sectors in each country to perform
better their duties, and ultimately provide the necessary background for the country's economic growth and
development. If the banks can do this, it is important to properly allocate financial resources to eligible
customers. Proper allocation of financial resources, while achieving the above objective, will provide the
necessary ground for the continued life of the banks. In this case, it is important to correctly identify the
risk-averse customers before granting them facilities in order to enhance the effectiveness of the decisions
taken. Estimating that a company will go bankrupt in the future is very important for facilitators and
creditors, so finding the model that best fits the companies has always been a concern.
Method: In this study, credit risk assessment of legal clients of Maskan Bank was investigated by using
logistic regression and feature selection by genetic algorithm. It is noteworthy that for the dataset, a
normalization method was used (ratio of distance from mean to data standard deviation) and the results
were obtained based on both normalized and abnormal data sets to determine the impact of data
normalization on the data set to get the right prediction percentage from customers.
Results: In this study, the regression coefficients on the Maskan Bank dataset are calculated based on the
logistic regression model using IVIOS software, and then the prediction of correct results based on the
logistic regression will be obtained in the MATLAB mathematical software. In addition, based on feature
selection with genetic algorithm, the results of logistic regression are optimized. Most of the work done in
this field by logistic regression did not have a prediction percentage above%80 and this prediction method
was lower than other prediction methods in the lower classes. In this study, we have obtained the correct
prediction percentage of %94.8 based on the use of appropriate features collected from customers in the
Maskan Bankk dataset by using feature selection through genetic algorithm.