🌳 Vietnam Deforestation Risk Logistic Model
An interpretable logistic-regression baseline for predicting forest-loss risk at grid-cell level in Gia Lai Province, Vietnam.
Model Details
- Architecture: Logistic Regression (scikit-learn
StandardScalernormalized). - 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:
0.931 - AP:
0.926 - F1:
0.881 - Capture@10%:
0.190
- AUC:
- Area Transfer (K’Bang → Mang Yang):
- AUC:
0.743 - Capture@10%:
0.170
- AUC:
Ethical Considerations & Intended Use
For proactively identifying high-risk deforestation zones to support local conservation patrols and reserve boundary allocation. This model must NOT be used to justify punitive commercial land claims or displacement of forest communities without on-the-ground human verification.
Source Code & Reproducibility
- Source Pipeline Code: Borino88/deforestation-risk-vietnam
- Interactive Explorer Space: MahdiFattahi/deforestation-risk-explorer
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support