Home-Credit-Default-Risk

The main goal of this project is to build an optimal ML model which will predict if a loan applicant will be able to repay his/her loan. In Phase 2, we extended our work from Phase 1 and implemented Feature Engineering where we consider potential features from other tables, did feature selection from the derived features, analysis of feature importances and implemented Hyper Parameter Tuning. In feature engineering, we derived 6 additional features and we achieved an improvement over our baseline model however some feature models are leading to over-fitting and thus in our future scope we aim to select the most important features having balanced data to avoid over-fitting. The most important feature relevant to our goal was the Credit Annuity ratio of the current application. We identified this by implementing feature importances on our model. After performing Hyper-parameter tuning on logistic regression(our best pipeline) we achieved the test accuracy of 92.46%. In Phase 3, we extended our work from Phase 2 and we performed a deep learning algorithm which will predict our goal of the project. The Deep Learning Algorithm used was the Multi-Layer Perceptron which is a kind of Artificial Neural network. The main goal of this phase was to implement MLP and visualize the training model on TensorBoard. We also identified Data Leakage in our project and their respective reasons. Another goal of this phase was to improve our results from Phase 2 and we were successful in doing that.