A deep learning model for predicting thyroid disease using patient demographics and blood test results. This project aims to assist healthcare providers with early intervention and improved disease management.
This project implements a multi-class neural network classifier to predict different categories of thyroid disease. The model achieves approximately 75% accuracy on the test set and handles class imbalance through weighted sampling.
- Data preprocessing pipeline for handling missing values and outliers
- Feature scaling and encoding for categorical variables
- Custom PyTorch Dataset implementation
- Deep neural network with batch normalization and dropout
- Class-weighted loss function to handle imbalanced data
- Learning rate scheduling and model checkpointing
- Python 3.8+
- PyTorch
- NumPy
- Pandas
- Scikit-learn
- Seaborn
- Matplotlib
- Handles missing values in thyroid test data
- Removes outliers and encodes categorical variables
- Scales numerical features
- Implements custom neural network for thyroid disease classification
- Uses multi-layer architecture with batch normalization and dropout
- Handles class imbalance with weighted loss
- Custom dataset class for data loading
- Training loop with learning rate scheduling
- Model evaluation and performance metrics
- Statistical analysis of thyroid metrics
- Distribution plots for key features
- Training/validation curves
- Classification performance metrics