Historical data were gathered from bank customers to determine whether a customer is a good or bad credit risk for a home equity loan. Bad risk customers are more likely to default on the loan.

Packages: pandas, numpy, sklearn, matplotlib, seaborn.

• First, I converted the target variable type to an object.
• I detected missing values

• I filled the missing values with average and most frequent values.

• I looked at the relationship of categorical variables with target variable.
• I grouped the target value according to the mean of the numerical values.Then I separated the different ones.
• Shapiro-wilks test was performed to check normality assumption.(All Non normal distributed).
• Then I grouped them and translated them into a categorical variable. Thus, I was able to test variables that we know to affect the target variable.
• I paid attention to the homogeneous distribution of the data and the meaning of the variable.
• I did a chi-square test on the variables I separated and all of them turned out to be related to the target variable.

If the number of delaying monthly payments is greater than 5, he / she is put in the bad risk group. The number of delaying payments per month is 5 and the value of properties of customers who are identified as good risk is different from others.

First, I transformed the categorical variables into dummy variables. I also split the data into train and tests sets with a test size of 20%.

I tried three different models and evaluated them using Accuracy and F1 score.

• I achieved 87% accuracy in the logistic regression model that I created with only debt-income ratio.

• So if we know only the debt-income ratio of a new customer, we can find out that this customer is either a good risk or a bad risk with 87% accuracy.

• Deptinc_score stands out in the feature importance chart.

• When examining the model performance, F-score (1) (bad risk) was considered. We want to find the bad risk client.
• XGB model outperformed the other approaches.