Data analytics and predictive modelling of credit risk of bank customers
To reduce losses and increase profits, financial organizations must evaluate credit risk. In this article, we propose an ensemble model for assessing credit risk that combines the methods of neural networks and different machine-learning models. The dataset used for this research is obtained from th...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/167777 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | To reduce losses and increase profits, financial organizations must evaluate credit risk. In this article, we propose an ensemble model for assessing credit risk that combines the methods of neural networks and different machine-learning models. The dataset used for this research is obtained from the IEEE DataPort and consists of various credit risk-related features. A comprehensive pipeline is created that includes feature decomposition with an auto encoder, data balancing with adaptive synthetic sampling (ADASYN) and random under sampling, feature removal using the interquartile range (IQR) method, and elimination of highly correlated features through Pearson correlation. Comparative analysis is done between the proposed ensemble model and standalone traditional models like Logistic Regression, SVC etc. The results show that the proposed ensemble model outperforms individual models in predicting credit risk in both balanced and original datasets. The ensemble model performs exceptionally well on both balanced and original dataset in accurately classifying instances, distinguishing between positive and negative cases, and selecting relevant samples, with a harmonized average of precision and recall. On the balanced and original dataset, the proposed ensemble model achieves an accuracy, precision, re-call and F1 score of 0.9995 and 0.997 respectively. In essence, the proposed ensemble model provides a very effective and dependable solution for credit risk evaluation, greatly outperforming individual models in prediction performance. Financial organizations can use this approach to better manage risks associated with credit lending. |
|---|