Session

Saurav Singla

Saurav Singla

Advisor/Senior Data Scientist

Mindrops

Machine Learning to Predict Credit Risk in Lending Industry

Tuesday,  Mar 23 | 11:00AM - 11:45AM US ET

Level: Beginner

Banks are a critical part of economic growth. Banks have credit risk like other businesses in the financial sector. Predicting credit risk is the main area of concern for the banking sector in most of the countries around the world. Credit risk arises when the borrower fails to pay the money borrowed. The lender used credit bureaus for consumer credit history, however rapid deviations in consumer behavior and market conditions have made credit bureau information unreliable for consumer creditworthiness. Objective: To improve the credit risk prediction model in the lending industry by building and comparing different supervised learning classifiers. Methods: Neural Network, Deep Neural Network, Support Vector Machine, Recursive Partitioning, Random Forest, Logistic Regression, Random Grid Search, Gradient Boosting Algorithm, Ensemble were used for prediction. Different techniques were further compared. Results: Decrease in False Negative Rate led to increasing in False Positive Rate and vice versa. GBM with the lowest FPR. Many thrilling subjects have appeared during work. One of them is the low performance of the Deep Neural Network which can be improved by further research. Conclusion: Based on the metrics and comparisons, we concluded that the methods built on artificial intelligence can improve the risk prediction of financial institutions. A bad loan classified as a good loan is very costly for the organization. Gradient Boosting Algorithm can be an appropriate alternative. GBM gives better results as compared to other algorithms. and GBM attains less volume of bad customers accepted. So, GBM attains less cost of misclassification.