This repository contains a Jupyter notebook titled Autoencoder.ipynb dedicated to implementing an autoencoder model for credit risk assessment using customer financial data. The notebook is structured to provide a comprehensive guide on data handling, feature engineering, and model training with detailed explanations and Python code.
- Dataset Overview: Includes financial transaction data and customer profiles with features like income, savings, debt levels, and categorical variables for personal attributes.
- Preprocessing Steps: Data cleaning, handling missing values, and feature normalization using pipelines for numerical and categorical data.
- Model Implementation: Uses an autoencoder model, ideal for dimensionality reduction and anomaly detection in dense datasets.
- Model Training and Validation: Detailed steps on training the model with a train-test split, hyperparameter tuning using grid search, and model evaluation.
This project aims to utilize deep learning techniques to predict credit risk and default probability, providing insights that can help in making informed financial decisions.