SecureHealthIoT

KO | EN

gitlite — search

#python #react #typescript #golang #android #machine-learning #rust #macos #java #pytorch #security #ios #openai #open-source

SecureHealthIoT

(★ 10)

Kaggle-based symptom-to-disease prediction pipeline with Gradio and Hugging Face deployment

SecureHealthIoT Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# SecureHealthIoT Disease Prediction

Production-style symptom-based disease prediction stack with:

- leakage-aware grouped splits
- external validation split
- out-of-domain validation on second independent dataset
- probability calibration (`CalibratedClassifierCV`)
- model versioning + reproducible Kaggle GPU training
- Hugging Face model + dataset + Space deployment
- user-friendly app with symptom mode + report/image-assisted mode

## Live Links

- Model: https://huggingface.co/ShiroOnigami23/securehealthiot-disease-model
- Dataset: https://huggingface.co/datasets/ShiroOnigami23/securehealthiot-disease-dataset
- App Space: https://huggingface.co/spaces/ShiroOnigami23/securehealthiot-disease-app

## Core Reliability Features

1. Grouped splitting by symptom signature to reduce leakage across splits.
2. Separate external validation set.
3. Calibration using sigmoid scaling for more reliable probabilities.
4. Leakage report in metrics (`overlap_signature_*` checks).
5. True OOD scoring using `kaushil268/disease-prediction-using-machine-learning`.
6. Versioned artifacts in `artifacts/model_registry`.

## App Features

1. Quick Symptom Predictor:
   - multi-select symptoms
   - top-3 predictions
   - abstain/reject mode with configurable threshold
   - confidence band + precaution summary
   - clinician-facing PDF report export
2. Report/Image Assisted Analysis:
   - upload `.txt`, `.csv`, or `.pdf` report
   - optional image upload with quality diagnostics
   - symptom extraction from report text + combined prediction
   - clinician-facing PDF report export
3. Preliminary supportive-care and prescription guidance with strict medical disclaimer.

## Install

```bash
pip install -r requirements.txt
```

## Local Training

```bash
python -m disease_ml.train --note "reliability_pass_v2"
```

To disable OOD validation:

```bash
python -m disease_ml.train --disable-ood-validation
```

Outputs:

- `artifacts/latest_model.joblib`
- `artifacts/model_registry/<timestamp>_<model>/model_bundle.joblib`
- `artifacts/model_registry/<timestamp>_<model>/metrics.json`

## Kaggle GPU Training

Push + run:

```bash
kaggle kernels push -p kaggle_kernel
python scripts/wait_kaggle_kernel.py
kaggle kernels output aryansingh21fd/securehealthiot-disease-trainer-v1 -p kaggle_pull
```

Sync outputs to HF:

```bash
set HF_TOKEN=YOUR_TOKEN
python scripts/upload_kaggle_outputs_to_hf.py --outputs-dir kaggle_pull
```

## Run App Locally

```bash
python app.py
```

Open:

- `http://localhost:7860`

## Publish/Refresh Hugging Face Space

```bash
set HF_TOKEN=YOUR_TOKEN
python scripts/publish_space.py --space-id ShiroOnigami23/securehealthiot-disease-app --model-repo-id ShiroOnigami23/securehealthiot-disease-model
```

## Medical Notice

This software is for educational and decision-support usage. It is not a medical device and not a replacement for licensed clinical judgment.
Any prescription decision must be made by a licensed physician after direct consultation.

SecureHealthIoT

SecureHealthIoT Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

SecureHealthIoT

SecureHealthIoT Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README