🧠 Stroke Prediction Model in R (Interactive Web App & Report)

Link Here and wait for 60s to Open Live App

"Data science in healthcare isn't just about prediction — it's about making insights useful, understandable, and actionable. This project brings that philosophy to life."

This project presents a complete, end-to-end solution for predicting stroke risk using real-world healthcare data. Developed entirely in R and powered by Shiny, it combines interactive user interfaces, advanced machine learning modeling, and visual communication — all integrated within a single, beautifully styled RMarkdown document.

🌐 Live App: Try the Stroke Predictor App
📄 Full Source File: Build-deploy-stroke-prediction-model-R.Rmd

---

🚀 Project Overview

The goal of this project is to empower healthcare professionals and researchers with a simple, interactive tool that estimates the likelihood of a stroke based on patient-specific attributes such as age, glucose levels, hypertension status, and more.

Using modern data science techniques, the app was designed to:

• Preprocess and explore real-world clinical data
• Train and evaluate multiple predictive models
• Visualize key risk factors and performance metrics
• Deploy an interactive prediction interface using R Shiny
• Serve as a reproducible and professional report via R Markdown

---

📦 App Features

✅ Interactive Web App (Shiny-based)
✅ Responsive Dashboard UI with bslib, custom CSS & watermark
✅ Real-time stroke risk prediction based on 10+ user inputs
✅ Machine Learning Models: Logistic Regression, Random Forest, XGBoost
✅ Integrated SMOTE balancing for handling imbalanced data
✅ Full EDA, preprocessing, model tuning, and evaluation
✅ Faceted density plots, ROC curves, variable importance, and calibration
✅ Fully reproducible in a single .Rmd file

---

📊 Data & Preprocessing

The model is trained on a public healthcare dataset (healthcare-dataset-stroke-data.csv) with 11 predictors and a binary outcome (stroke).

Key preprocessing steps included:

• Handling missing values (PMM imputation via mice)
• Conversion and normalization of categorical/numeric fields
• Addressing class imbalance using SMOTE from the themis package
• Feature engineering using the recipes package

---

🧠 Model Building

Using the tidymodels framework, we trained and evaluated:

• Logistic Regression — for interpretability
• Random Forest — for robustness and nonlinearity
• XGBoost — for gradient-boosted performance

Metrics used: AUC, Accuracy, Recall, F1, and Precision

Cross-validation was performed using stratified 5-fold CV.

---

🎨 Shiny App UI Highlights

• Custom blue-themed UI using bslib::bs_theme()
• Brain watermark background (brain_watermark.png)
• Predict button with rounded hover effect
• Two-column result boxes (green/red) showing stroke vs. no-stroke probabilities
• Uses renderUI() and observeEvent() for interactivity

---

📁 File Structure

File	Description
Build-deploy-stroke-prediction-model-R.Rmd	Full report + shiny app + code (main file)
healthcare-dataset-stroke-data.csv	Dataset used for training the models
brain_watermark.png	Watermark image for UI background
README.md	This file

---

🧪 How to Run This Project Locally

1. Clone the repo:

   git clone https://github.com/your-username/stroke-prediction-shiny.git
   cd stroke-prediction-shiny

2. Open Build-deploy-stroke-prediction-model-R.Rmd in RStudio

3. Click "Run Document" or run this in console:

---

📤 Deployment

This app is deployed on shinyapps.io, enabling real-time, browser-based interaction without needing R installed.

Visit the live app: 🌐 https://mashhood23.shinyapps.io/Build-Deploy-Stroke-Prediction-Model-R/

---

📚 Author

Mashhood Raza Khan

📧 mashhood1223@gmail.com

📍 India

---