The Goal of this project was to predict the Risk factor of loans approved by the German Banks. For this, we had historical data of loans provided by the banks, and cleaned and preprocessed data to make it useful for our prediction. Deployed, the classification algorithm such as Logical Regression with(l1 & l2) and without Regression, K nearest Neighbour method, Random Forest, Gradient Boosting, and finally stacking of Gradient Boosting with Logical regression and KNN. From the analysis of evaluation matrics of these models, we found that logical regression and Stacking of KNN with Gradient Boosting performed well, with around 71% accuracy.
Also tried to predict using pycaret open source package for a low code project.
The folder contains:
- Jupyter notebook containing python code for the analysis and modelling(https://github.com/vidhya-5684/German-Credit-Risk/blob/main/German_Credit_Risk.ipynb)
- Jupyter notebook for modelling with pycaret(https://github.com/vidhya-5684/German-Credit-Risk/blob/main/creditcard_pycaret.ipynb)
- A python code script 'colors.py' used for plotting of Confusion Matrix.