README.md

metadata

title: Stroke Prediction Model
emoji: 🧠
colorFrom: red
colorTo: blue
sdk: docker
app_file: app.py
pinned: false

Stroke Risk Prediction Model

This Hugging Face Space hosts a machine learning model for stroke risk prediction based on various health and demographic factors.

Model Overview

• Type: Random Forest Classifier
• Target: Binary classification (stroke risk)
• Output: Probability of stroke risk + risk category
• Features: Age, gender, hypertension, heart disease, BMI, glucose levels, and more

API Endpoints

POST /api/predict

Predicts stroke risk based on patient health data.

Request Format

{
  "gender": "Male",
  "age": 67,
  "hypertension": 1,
  "heart_disease": 0,
  "ever_married": "Yes",
  "work_type": "Private",
  "Residence_type": "Urban",
  "avg_glucose_level": 228.69,
  "bmi": 36.6,
  "smoking_status": "formerly smoked"
}

Response Format

{
  "probability": 0.72,
  "prediction": "High Risk",
  "stroke_prediction": 1,
  "risk_factors": [
    "Advanced Age (>65)",
    "Very High Blood Glucose (>200)",
    "Obesity (BMI > 30)",
    "Former Smoker"
  ],
  "important_features": [
    {"feature": "avg_glucose_level", "importance": 0.31},
    {"feature": "age", "importance": 0.28},
    {"feature": "bmi", "importance": 0.15}
  ],
  "execution_time_ms": 50,
  "using_model": true,
  "model_version": "1.0"
}

GET /

Returns information about the model and API usage.

Parameter Details

Parameter	Description	Type	Values
gender	Gender of the patient	String	"Male", "Female", "Other"
age	Age in years	Number	0-120
hypertension	Whether patient has hypertension	Number	0 (No), 1 (Yes)
heart_disease	Whether patient has heart disease	Number	0 (No), 1 (Yes)
ever_married	Whether patient has ever been married	String	"Yes", "No"
work_type	Type of work/employment	String	"Private", "Self-employed", "Govt_job", "children", "Never_worked"
Residence_type	Type of residence	String	"Urban", "Rural"
avg_glucose_level	Average glucose level in blood (mg/dL)	Number	50-300+
bmi	Body Mass Index	Number	10-50+
smoking_status	Smoking status of patient	String	"never smoked", "formerly smoked", "smokes", "Unknown"

Risk Categories

• Very Low Risk: < 10% probability
• Low Risk: 10-20% probability
• Moderate Risk: 20-40% probability
• High Risk: 40-60% probability
• Very High Risk: > 60% probability

Model Training

This model was trained on a dataset of over 5,000 patients with various health metrics and stroke outcomes. It identifies key risk factors like:

• Age
• High blood pressure
• Heart disease
• Glucose levels
• BMI
• Smoking status

Integration with BrainWise App

This model serves as the backend for the BrainWise app's stroke risk calculator, providing users with risk assessments based on their health metrics.

Disclaimer

This model is for informational purposes only and should not replace professional medical advice. Always consult with healthcare providers for medical decisions.