🦠 Breast Cancer Classification with K-NN

📋 Overview

Machine learning model for breast cancer classification using K-Nearest Neighbors (K-NN) algorithm. Built with scikit-learn on the Wisconsin Breast Cancer dataset, achieving 96%+ accuracy in distinguishing malignant from benign tumors.

✨ Key Features

• High Accuracy: 96% classification accuracy on test data
• Feature Engineering: Optimized cell characteristic selection
• Cross-Validation: K-fold validation for robust performance
• Hyperparameter Tuning: Grid search for optimal K value
• Visualization: Confusion matrix and feature importance plots
• Jupyter Notebook: Interactive analysis and results

📊 Model Performance

Metric	Score
Accuracy	96.0%
Precision	95.8%
Recall	96.3%
F1-Score	96.0%
AUC-ROC	0.973

Confusion Matrix

                Predicted
              Benign  Malignant
Actual Benign    71        2
      Malignant   3       38

🚀 Getting Started

Installation

git clone https://github.com/masoud-rafiee/breast-cancer-analysis-knn.git
cd breast-cancer-analysis-knn
pip install -r requirements.txt

Running the Analysis

# Run Jupyter notebook
jupyter notebook breast_cancer_knn.ipynb

# Or run Python script
python train_model.py

📊 Dataset

Wisconsin Breast Cancer Dataset (Built-in to scikit-learn)

• Samples: 569 cell samples
• Features: 30 numeric features
• Classes: Malignant (212) | Benign (357)
• Features include: radius, texture, perimeter, area, smoothness, compactness, etc.

🎯 K-NN Algorithm Details

How It Works

1. Training: Store all training samples
2. Prediction: For new sample, find K nearest neighbors
3. Classification: Majority vote determines class
4. Distance Metric: Euclidean distance in feature space

Optimal Parameters

• K (neighbors): 7 (determined via cross-validation)
• Distance Metric: Euclidean
• Weights: Distance-weighted voting
• Feature Scaling: StandardScaler normalization

🔬 Key Findings

1. Most Important Features:

   • Worst concave points
   • Worst perimeter
   • Mean concave points

2. K Value Impact: K=7 provides optimal bias-variance trade-off

3. Feature Scaling: Critical for K-NN performance (15% accuracy boost)

📊 Visualizations

The notebook includes:

• Feature distribution plots
• Correlation heatmap
• Decision boundary visualization (2D PCA)
• K value vs accuracy curve
• Confusion matrix
• ROC curve

🛠️ Technologies

• Python 3.8+: Core language
• scikit-learn: Machine learning framework
• pandas: Data manipulation
• NumPy: Numerical computing
• Matplotlib/Seaborn: Visualization
• Jupyter: Interactive analysis

📝 Project Structure

breast-cancer-analysis-knn/
├── breast_cancer_knn.ipynb    # Main analysis notebook
├── train_model.py             # Training script
├── requirements.txt
├── models/
│   └── knn_model.pkl          # Trained model
├── plots/                     # Generated visualizations
└── README.md

⚠️ Medical Disclaimer

This is an educational project. NOT for clinical diagnosis. Always consult medical professionals for health decisions.

📝 License

MIT License

👤 Author

Masoud Rafiee

• GitHub: @masoud-rafiee
• LinkedIn: masoud-rafiee

🙏 Acknowledgments

• CS331 - Machine Learning
• Bishop's University
• UCI Machine Learning Repository

📚 Further Reading

• K-NN Algorithm Overview
• Wisconsin Breast Cancer Dataset

---

Machine learning for healthcare 🦠🔬