Multiple health-related regression neural network models trained accessible securely on the internet.
The datasets were gathered from UC Irvine Machine Learning Repository:
- Heart Disease
- Maternal Health Risk
- Estimation of Obesity Levels Based On Eating Habits and Physical Condition
- CDC Diabetes Health Indicators
This project was developed in Python3 with a virtual environment through venv. You can find and install the dependencies by utilizing the requirements.txt provided.
- The models were created using PyTorch. There's one notebook for each model. These notebooks may be used to train or retrain the models as needed. A minimal setup was used when programming on these notebooks. Other than PyTorch, the only core libraries necessary are numpy and pandas. For model export and import, the corresponding onnx library or runtime is required.
Inside the /onnx/ directory, you will find the exported models in .onnx format. Please open their corresponding notebooks to see the model architecture, or open these .onnx files on Netron.
The web application has its own /onnx/ directory which is a copy of the /onnx on the project root. The only requirement when hosting the web application is the ability to securely read resources. Therefore opening the web application's build files require a protocol other than the file protocol which is insecure. All model computations happen on the client-side (i.e. the device running the web application) for privacy and security. Please open the /web/ directory for more information.