MahdiFattahi/vietnam-deforestation-risk-random-forest

🌳 Vietnam Deforestation Risk Random Forest Model

A high-performance Random Forest classifier predicting forest-loss risk at 1 km grid-cell resolution.

Model Details

• —Architecture: Random Forest (nestimators=400, maxdepth=22, minsamplesleaf=2).
• —Inputs: 1 km aggregated raster features (elevation, slope, precipitation, historical forest loss, baseline canopy).
• —Target: Binary deforestation indicator.

Performance Metrics (Manuscript Table 3)

• —Provided Split (Out-of-sample):
• —AUC: 0.942
• —AP: 0.914
• —F1: 0.878
• —Capture@10%: 0.184
• —Area Transfer (K’Bang → Mang Yang):
• —AUC: 0.767
• —Capture@10%: 0.145

Ethical Considerations & Limitations

For proactive conservation patrol planning only. Do not use for automated land-rights decisions.