Abstract
The major objective of this project is to develop a credit scoring model with reliable accuracy and competitive speed. We build the credit scoring system based on neural network model, and compare it with other methods including traditional linear regression and logistic regression. In addition, we also test the ability to optimize performance of two feature selection models.
Researchers:
Research Group (2015 Fall):
Yang Liu, Master in Financial Engineering, Graduated in Jan 2016
Yang Li, Master in Financial Engineering, Graduated in Jan 2016
Haoshuo Zhang, Master in Financial Engineering, Graduated in Jan 2016
Advisor:
Dr. Rupak Chatterjee
Dr. Ionut Florescu
Research Topics:
Neural Network, Feature Selection, Logistic Regression, Credit Scoring System
Main Results:
To improve the accuracy, we apply the feature selection technique, which improves our accuracy by 10%. Our model allows us to forecast a rough credit rating situation (within one rating level error) with a confidence level of 95%, and an accurate credit rating with a confidence level around 65% to 70%.
Conclusions
At the end of the project, we are able to predict the credit rating of any firm with public financial data using the neural network model we built. A high confidence level is achieved when we only want a rough estimation of the credit rating situation. To sum up, our model works.