The work presented here was made by Thúllio Debortoli Moreira Zanetti under the guidance of Professor Frederico Gadelha Guimarães.
This Undergraduate thesis makes a comparative approach of three different Artificial Intelligence techniques to solve the credit scoring classification problem. The models presented here are Decision Tree, Random Forest and Multilayer Perceptron. Credit analysis is a very popular topic among Artificial Intelligence and Machine Learning researchers. The classification models obtained are compared by their acuracy, AUC (Area Under the ROC Curve), type I error rate and the model’s explainability, bearing in mind the importance of credit rejection or approval analysis to credit institutions.
The data cleaning methodology, the steps of models trainning and the results obtained are presented on the thesis and all the code used is in the IPython Notebook "TCC2_II-Code.ipynb".