This repository contains the code and resources for the Home Credit Default Risk project. The project aims to predict the likelihood of default for loan applicants using machine learning techniques as well as the credit amount prediction on potential default loan applicants.
The Home Credit Default Risk project focuses on predicting whether a loan applicant is likely to default on their loan payment. The analysis is based on various features and characteristics of the applicants, such as their credit history, income, and demographic information. By accurately predicting default risk, financial institutions can make better decisions regarding loan approvals and risk management. This repository provides the necessary code and resources to replicate the analysis and predictions. The main code file is provided in both Jupyter Notebook (ipynb) and HTML format.
The dataset used in this project is available from the Home Credit Default Risk Kaggle competition. It consists of anonymized data containing information about loan applicants, including their financial and credit history. The dataset is split into training and testing subsets. To reproduce the analysis and predictions, you will need to download the dataset from Kaggle and place it in the appropriate folder. Please refer to the Kaggle competition page for more details on the dataset and its features.
Download the dataset from the Home Credit Default Risk Kaggle competition and place it in the data folder. Clone this repository to your local machine or download the ZIP file. Ensure that the image-folder directory, containing relevant images, is present in the same folder as the Jupyter Notebook and HTML files. Launch Jupyter Notebook and open the home_credit_default_risk.ipynb file. Run the notebook cells to execute the code and reproduce the analysis. Alternatively, you can also view the HTML files for a static version of the notebook.
The project aims to predict the likelihood of default for loan applicants. The analysis includes data preprocessing, feature engineering, model training, and evaluation. Detailed results and findings can be found in the Jupyter Notebook or HTML file. The final model targets are as follow.
1_ML_classification_models_comparison.ipynb
-> Multiple classification models comparison based on AUC ROC score
2_Light_GBM_classification_model_prediction_and_improvement
-> An AUC ROC score improvement based on feature engineering and hyperparameter tuning
3_Light_GBM_regression_model_prediction
-> Credit amount prediction based on Light GBM regression model