Diabetes Indicator - Backend

API for the Diabetes Indicator App

• Acess the website here: https://kmads.dev/diabetes-indicator
  
• Acess the repository here: https://kmads.dev/source-code/diabetes-indicator

Deployment

• Frontend: Source | Live Deploy
• Backend + Machine Learning: Source

API

You can acess the API here:

https://api.kmads.dev/diabetes-indicator/<args>

Endpoints:

Endpoint	Method	Description
/	GET	API status with current model info
/health	GET	Health check endpoint
/models	GET	List all available trained models
/predict	GET	Diabetes prediction with confidence percentage

Health Indicators (Prediction Input)

The /predict endpoint accepts 17 comma-separated health indicators:

#	Field	Type	Values
1	HighBP	int	0, 1
2	HighChol	int	0, 1
3	CholCheck	int	0, 1
4	BMI	float	0-275
5	Smoker	int	0, 1
6	Stroke	int	0, 1
7	HeartDiseaseorAttack	int	0, 1
8	PhysActivity	int	0, 1
9	Fruits	int	0, 1
10	Veggies	int	0, 1
11	HvyAlcoholConsump	int	0, 1
12	AnyHealthCare	int	0, 1
13	NoDocbcCost	int	0, 1
14	GenHlth	int	1-5
15	DiffWalk	int	0, 1
16	Sex	int	0, 1
17	Age	int	0-13 (age groups)

Architecture

    diabetes-indicator-backend
    ⊢ trained-models/ 
    |  ⌞ <.pkl files> (models trained and ready for production)
    ⊢ machine-learning/
    |  ⌞ ml.ipynb (python notebook where the models are trained)
    ⊢ main.py (API made with fastapi and uvicorn)
    ⊢ requirements.txt (python dependencies)
    ⊢ vercel.json (vercel deploy config)
    ⌞ README.md

Flow

• client Request request prediction from the frontend
• frontend gets prediction from api/
• backend choose the latest model (.pkl) from trained-models/
• trained-models/ receives the model (.pkl) from machine-learning/

Simplified:

• Client -> Frontend <-> Backend -> <model>.pkl (trainedModels/)

Copyright

No restrictions of use, just give some credits if you want