The application of machine learning algorithms in credit scoring

Ing. Niguss Solomon Hailegnaw

Master's thesis

Abstract:

In this thesis the principle of machine learning by using logistic regression, decision tree and neural network algorithms have been used for the purpose of credit scoring. To achieve this, a big data set of borrowers with more than 8000, 000 observations from an American lending company called LendingClub have been downloaded. The selection of important features or variables was done by selecting variables having missing values with less than 50% of the total observations. Missing values was were detected, filled with mean, mode or median of the variables and or removed from the dataset accordingly. The data was also checked for any duplicated rows, outliers and collinearity. After the preprocessing for the data some data visualization has been done for the detection of some possible relation between the variables and the target variable called loan_status. After all the credits rating was modeled by using logistic regression, decision trees and neural networks. Their prediction quality checked by using confusion metrics criteria and Receiver Operator Characteristic (ROC) areas under curve (AUC). Based on the outcome of our modeling the Logistic regression overtakes both decision tree and neural network. Finally, based on the output of this work the use of logistic regression for the specific case of credit rating is recommended than decision tree and neural network. Loan amount, loan term, annual income, borrower’s total debt payment over obligation, inquiries last 6 months, revolving utilities, total account and last fico range high are the most important power full predictor of defaulters p value of 0.000.

Keywords

Machine learning, ., credit rating, logistic regression, decision trees, neural networks

Language used: English

Date on which the thesis was submitted / produced: 31. 3. 2020

Thesis defence

• Supervisor: Ing. Tomáš Hlavsa, Ph.D.

Czech University of Life Sciences Prague

Faculty of Economics and Management

Master programme / field:
Systems Engineering and Informatics / Informatics