Apppublic

MotavalD/Stroke_Prediction

sourceHugging Faceupdated 3mo agoView on Hugging Face

0likes

App README

🫀 Stroke Risk Prediction API

A machine learning API that predicts stroke risk from patient health data. Built with FastAPI, Scikit-learn, and Docker — deployed as a production-ready REST endpoint.

What it does

Takes patient health indicators as input and returns a binary stroke risk prediction (0 = low risk, 1 = high risk). The model was trained on a healthcare stroke dataset with preprocessing, feature engineering, and class-imbalance handling.

Tech stack

Layer	Tool
API framework	FastAPI
ML model	Scikit-learn (best-performing classifier)
Preprocessing	Pandas, NumPy, StandardScaler
Serialization	Pickle
Deployment	Docker

API usage

Endpoint: POST /predict

Request body (JSON):

json

{
  "age": 67.0,
  "hypertension": 0,
  "heart_disease": 1,
  "avg_glucose_level": 228.69,
  "bmi": 36.6,
  "gender_Male": 1,
  "gender_Other": 0,
  "ever_married_Yes": 1,
  "work_type_Never_worked": 0,
  "work_type_Private": 1,
  "work_type_Self_employed": 0,
  "work_type_children": 0,
  "Residence_type_Urban": 0,
  "smoking_status_formerly_smoked": 1,
  "smoking_status_never_smoked": 0,
  "smoking_status_smokes": 0
}

Response:

json

{
  "prediction": 1
}

1 = high stroke risk, 0 = low stroke risk

Input features explained

Feature	Description
`age`	Patient age in years
`hypertension`	1 if patient has hypertension
`heart_disease`	1 if patient has heart disease
`avg_glucose_level`	Average blood glucose level (mg/dL)
`bmi`	Body mass index
`gender_Male` / `gender_Other`	One-hot encoded gender
`ever_married_Yes`	1 if ever married
`work_type_*`	One-hot encoded work category
`Residence_type_Urban`	1 if urban resident
`smoking_status_*`	One-hot encoded smoking history

Run locally

bash

# Clone the space
git clone https://huggingface.co/spaces/MotavalD/Stroke_Prediction

# Build and run with Docker
docker build -t stroke-api .
docker run -p 7860:7860 stroke-api

# Test the API
curl -X POST http://localhost:7860/predict \
  -H "Content-Type: application/json" \
  -d '{"age": 67, "hypertension": 0, "heart_disease": 1, ...}'

GitHub

Full source code and model training notebook available at: 👉 github.com/dhruv-motaval

Built by [Dhruv Motaval](https://www.linkedin.com/in/dhruv-motaval/) — B.Tech AI, Parul University