Machine Learning Models for Bad Loans Prediction

In this project we try to predict bad loans with Machine Learning models. We approach the problem as a Supervised Classification problem. We conduct a preprocessing step, and deal with imbalance. As a first step we apply various Machine Learning models and identify the best ones. With the remaining we try to improve the accuracy of these models with Hyperparameter Optimization and Cross Validation. Our findings suggest that these models are extremely good in predicting Bad Loans when they occur, however we see a relatively high value for False Positives. The focus of our research is in trying to reduce the False Positives for this problem. We try to infer were the problem is coming from (Oversampling techniques, preprocessing). Neither of those have nothing to do so we approach this problem with an ensemble model combining the most appropriate models to reduce False Positives