🧠 Stroke Classification - Spring 2025 Kaggle Competition

This repository contains my solution for the Spring 2025 Classification Competition hosted on Kaggle. The objective is to predict the risk of stroke in patients based on clinical and demographic data using robust machine learning models.

---

🏆 Competition Goal

Predict whether a patient is likely to experience a stroke using provided health-related features. The classification task is evaluated based on F1 Score and ROC-AUC.

---

📁 Dataset Overview

• Train Set: Labeled clinical and demographic features.
• Test Set: Unlabeled data for predictions.
• Target Variable: stroke (1 = stroke, 0 = no stroke)
• Features:
  • gender, age, hypertension, heart_disease, ever_married, work_type,
    Residence_type, avg_glucose_level, bmi, smoking_status

---

🧪 Evaluation Metric

• Primary Metric: F1 Score
• Secondary Metric: ROC-AUC Score

---

📊 Model Performance (Cross-Validation)

Model	F1 Score	ROC-AUC Score
🥇 Stacking	0.973	0.997
🥈 XGBoost	0.970	0.997
🥉 LightGBM	0.968	0.996
Random Forest	0.967	0.996
Decision Tree	0.903	0.961
K-Nearest Neighbors	0.734	0.813
SVM	0.664	0.665
Logistic Regression	0.530	0.634
Naive Bayes	0.324	0.666

---

📂 Repository Structure

stroke_detection_classification/
│
├── data/
│ ├── raw/ # Original raw datasets
│ └── processed/ # Cleaned & engineered datasets
├── notebooks/ # Jupyter notebooks (EDA, modeling)
├── reports/ # Model predictions, evaluation reports
├── src/ # Core Python code for modeling & preprocessing
│ ├── preprocessing.py
│ ├── modeling.py
│ ├── evaluation.py
│ └── utils.py
│ ├── models/ # Trained models (.pkl)
├── app.py # Streamlit web application
├── requirements.in # Dependency list (editable)
├── requirements.txt # Locked dependencies
└── README.md

---

🚀 How to Run the Project

1. Clone the repository
   git clone https://github.com/hemant-rout/stroke_detection_classification.git
cd stroke_detection_classification

2. 🔧 Setup Environment

   2.1 Install pip-tools if not already installed
   pip install pip-tools

   2.2 Compile and install dependencies
   pip-compile requirements.in
pip install -r requirements.txt

3. ▶️ Run the Streamlit App
   a. Local Machine:
   streamlit run app.py
   or
   b. Streamlit hosted app
   https://strokedetectionclassification-yt5ojuyg6orf2nsxs85ab8.streamlit.app/

---

🧪 Try It Yourself (Sample Inputs) Use the following examples in the app's sidebar to see prediction results:

gender age hypertension heart_disease ever_married work_type Residence_type avg_glucose_level bmi smoking_status

Male 68 0 0 Yes Private Rural 95.82 27.5 never smoked
Female 42 0 0 Yes Private Rural 84.18 26.1 smokes

🧠 The app will predict whether the patient is at risk of stroke and show the probability score.

---

📌 Future Improvements
🔧 Advanced Hyperparameter Tuning

🧬 Feature Selection & Dimensionality Reduction

• 🧠 Model Interpretation with SHAP
• 🔁 Automated Retraining Pipeline
• 📈 Live Model Monitoring Dashboard

---

📚 References

• Kaggle Competition Page
• Streamlit Documentation
• scikit-learn Documentation
• [XGBoost & LightGBM Docs](https://xgboost.readthedocs.io/, https://lightgbm.readthedocs.io/)

---

🧑‍💻 Author Hemant Kumar Rout 🔗 GitHub

Connect with me