Credit Scoring Prediction Using Deep Learning Models in the Financial Sector.

This project focuses on predicting credit scores using deep learning models in the financial sector. A hybrid framework is designed by combining three powerful deep learning algorithms: Wide & Deep DNN, Autoencoder + DNN, and TabNet. These models are employed to predict the creditworthiness of individuals by analyzing various financial features. The Wide & Deep DNN model captures both low-level and high-level patterns in the data, while the Autoencoder + DNN model helps in reducing dimensionality and extracting relevant features. TabNet, a more recent model, is leveraged for its ability to handle tabular data effectively, achieving the highest accuracy of 87.2%. The dataset, sourced from Kaggle, contains comprehensive information on financial attributes, and the models are evaluated based on their prediction accuracy. A Flask-based web application is developed to facilitate user interaction, including modules for Home, Register, Login, Prediction, and Logout. The integration of these models offers a robust solution for credit scoring, ensuring efficient feature representation and enhancing prediction accuracy.

Keywords: Credit Scoring, Deep Learning, TabNet, Autoencoder, Wide & Deep DNN, Financial Sector, Flask, Kaggle Dataset