README.md

metadata

title: AIO2025M03 AdaBoost Demo
emoji: 🚀
colorFrom: indigo
colorTo: blue
sdk: gradio
sdk_version: 5.38.0
app_file: app.py
pinned: false
license: mit

AIO2025 Module 03 - AdaBoost Demo

This interactive demo showcases AdaBoost (Adaptive Boosting) algorithms for both classification and regression tasks. The application provides a comprehensive interface for exploring sequential ensemble learning where each weak learner is trained adaptively based on the errors of previous learners through dynamic visualizations and real-time parameter adjustment.

🚀 Features

Core Functionality

• Dual Problem Types: Support for both classification and regression tasks
• Multiple Datasets: Built-in sample datasets including Titanic dataset
• Custom Data: Upload your own CSV/Excel files
• Interactive Parameters: Adjust boosting parameters in real-time
• Dynamic Input Generation: Automatic feature input creation based on dataset

AdaBoost Parameters

• Number of Estimators: Control sequential learning steps (limited to 1000 for performance)
• Learning Rate: Step size shrinkage for adaptive learning (0.0001-2.0)
• Max Depth: Individual weak learner depth (default: 1, decision stumps work best)
• Base Estimator: Decision tree with limited depth (weak learner)

Visualizations

• Boosting Progress Chart: Shows how predictions evolve sequentially as each estimator adapts to errors
• Individual Estimator Visualization: Detailed view of selected weak learner structure
• Feature Importance: Displays which features matter most across all estimators
• AdaBoost Process: Sequential aggregation display showing how predictions build up adaptively

🚀 Quick Start

1. Select Data: Choose from sample datasets or upload your own
2. Configure Target: Select the target column for prediction
3. Set Parameters: Adjust AdaBoost hyperparameters
4. Enter Features: Provide values for the new data point
5. Run Prediction: Execute the AdaBoost and view results

📊 Sample Datasets

Classification Datasets

• Titanic: Passenger survival prediction (default dataset)
• Iris: Classic 3-class flower classification
• Wine: Wine type classification based on chemical properties
• Breast Cancer: Binary classification for cancer detection

Regression Dataset

• Diabetes: Medical regression dataset

🛠️ Technical Details

Dependencies

• scikit-learn: AdaBoost implementation
• pandas: Data manipulation
• numpy: Numerical operations
• plotly: Interactive visualizations
• gradio: Web interface

Architecture

• Modular Design: Separated core logic in src/adaboost_core.py
• Dynamic UI: Automatic parameter and input generation
• Error Handling: Comprehensive validation and error messages
• Responsive Design: Adapts to different screen sizes

💡 Key Concepts

AdaBoost Benefits

• Sequential Learning: Each weak learner adapts to errors of previous learners
• Strong Performance: Often achieves high accuracy with simple weak learners
• Feature Importance: Robust importance scores through adaptive learning
• Adaptive Weighting: Focuses on misclassified examples by increasing their weights
• Error Reduction: Each iteration reduces prediction errors adaptively

Sequential Ensemble Learning

• Adaptive Reweighting: Increases weights of misclassified examples
• Weighted Voting: Final prediction is weighted sum of weak learner predictions
• Weak Learners: Uses simple estimators like decision stumps
• Error-Focused Training: Each new learner focuses on previously misclassified examples

🔧 Customization

Adding New Datasets

1. Place CSV files in the data/ directory
2. Update SAMPLE_DATA_CONFIG in app.py
3. Ensure target column is properly configured

Modifying Parameters

• Edit parameter ranges in the UI components
• Adjust default values for different use cases
• Add new parameter types as needed

📈 Performance Tips

• Number of Estimators: Limited to 1000 for optimal performance in this demo
• Learning Rate: Default 1.0 works well; lower rates (0.0001-0.1) create more conservative models, higher rates (1.0-2.0) for faster learning
• Max Depth: Decision stumps (depth 1) typically optimal for AdaBoost
• Weak Learners: Simple estimators work best to avoid overfitting

🎯 Use Cases

Classification

• Medical diagnosis (Titanic survival prediction)
• Customer segmentation
• Fraud detection
• Text classification

Regression

• Price prediction
• Demand forecasting
• Quality assessment
• Risk modeling

📝 Notes

• Auto-detection: Problem type automatically detected from target column
• Data Validation: Comprehensive input validation and error handling
• Memory Efficient: Optimized for sequential learning
• Real-time Updates: Instant parameter adjustment and visualization
• Estimator Selection: Interactive dropdown to explore individual weak learners (up to 100)
• Adaptive Nature: Each estimator adapts to previous learners' mistakes through reweighting

🔗 Related Resources

• Scikit-learn AdaBoost Documentation
• AdaBoost Explained
• AdaBoost Algorithm

---

This demo is part of AIO2025 Module 03 - Machine Learning Fundamentals