FedHealthDP Project

Project Overview

The FedHealthDP Project aims to train disease prediction models using federated learning with differential privacy. This approach leverages data from the MIMIC-III and Kaggle databases to create robust and privacy-preserving predictive models for cancer risk assessment.

Main Functional Modules

1. Data Preprocessing (scripts/preprocess.py):

   • Load data from sources
   • Handle missing values, normalization, and feature engineering

2. Model Definition (models/model.py):

   • Define deep learning model structure
   • Support federated learning and differential privacy architecture

3. Differential Privacy Optimizer (models/dp_optimizer.py):

   • Implement differential privacy optimization algorithms
   • Ensure data privacy during training

4. Federated Learning Implementation (scripts/federated_learning.py):

   • Core logic for federated learning
   • Manage model updates and aggregation across multiple clients

5. Training Script (scripts/train.py):

   • Train models, including local and federated training
   • Support configurable training parameters

6. Evaluation Script (scripts/evaluate.py):

   • Evaluate model performance
   • Generate reports and visualization results

7. Utility Functions (utils/data_utils.py and utils/eval_utils.py):

   • Common data processing and evaluation functions for use by other modules

Directory Structure

FedHealthDP_Project/
│
├── README.md              # Project introduction and instructions
├── config.yaml            # Configuration file
├── requirements.txt       # List of dependencies
├── main.py                # Project entry point
│
├── data/                  # Data directory
│   ├── kaggle/            # Kaggle datasets
│   └── mimiciii/          # MIMIC-III datasets
│
├── models/                # Models directory
│   ├── __init__.py        # Initialization file
│   ├── dp_optimizer.py    # Differential privacy optimizer implementation
│   └── model.py           # Model definition
│
├── scripts/               # Scripts directory
│   ├── preprocess.py      # Data preprocessing script
│   ├── train.py           # Training script
│   ├── federated_learning.py # Federated learning implementation script
│   └── evaluate.py        # Model evaluation script
│
├── test/                  # Test directory
│   ├── test_evaluate.py   # Evaluation tests
│   ├── test_federated.py  # Federated learning tests
│   ├── test_preprocess.py # Preprocessing tests
│   └── test_train.py      # Training tests
│
└── utils/                 # Utilities directory
    ├── __init__.py        # Initialization file
    ├── data_utils.py      # Data processing utilities
    └── eval_utils.py      # Evaluation utilities

## Datasets 
https://data.world/deviramanan2016/nki-breast-cancer-data/workspace/file?filename=NKI_cleaned.csv
/data/nci/breast_cancer.csv

https://www.kaggle.com/datasets/yasserh/breast-cancer-dataset
/data/kaggle/breast_cancer_0.csv

https://www.kaggle.com/datasets/reihanenamdari/breast-cancer
/data/kaggle/breast_cancer_1.csv

https://www.kaggle.com/datasets/raghadalharbi/breast-cancer-gene-expression-profiles-metabric
/data/kaggle/breast_cancer_2.csv