Forest Cover Type Classification

This project predicts the type of forest cover using cartographic and environmental features from the Covertype dataset (UCI). By applying tree-based multi-class classification models, it demonstrates how data-driven approaches can support ecological research and resource management.

Features

• Explored dataset distribution and target class balance
• Performed data cleaning and preprocessing (including categorical handling)
• Trained Random Forest and XGBoost models
• Evaluated models using Accuracy, Confusion Matrix, and Feature Importance
• Compared Random Forest vs XGBoost
• Performed hyperparameter tuning for improved performance

Dataset

• Source: UCI Covertype Dataset
• Features: Cartographic and environmental attributes (elevation, slope, soil type, aspect, etc.)
• Target: Forest cover type (7 classes)

Models & Results

Model	Accuracy	Precision	Recall	F1-score	Notes
Random Forest (untuned)	0.9551	0.96	0.96	0.95	Strong baseline
Random Forest (tuned)	0.9550	0.96	0.96	0.95	Minimal gain after tuning
XGBoost (untuned)	0.9533	0.95	0.95	0.95	Strong initial performance
XGBoost (tuned)	0.9601	0.96	0.96	0.96	Best model with ~96% accuracy

✅ XGBoost (tuned) achieved the best performance overall, with the most balanced precision–recall across all forest cover types.

Run the Project

git clone https://github.com/Minahil-Abid/ForestCoverType-Multi-Class-Classification.git
cd ForestCoverType-Multi-Class-Classification
pip install -r requirements.txt
jupyter notebook Forest_CoverType_Classification.ipynb

License

MIT License – free to use and share.