Feature Importance and Model Stability Data for Machine Learning-Based Mangrove Mapping in Sumbawa Regency, Indonesia (2025)

This dataset contains feature importance scores and model stability results from a multi-sensor machine learning mangrove mapping study in Sumbawa Regency, West Nusa Tenggara, Indonesia. Files: 1. RF_importance.csv - Random Forest Gini importance scores for 15 spectral predictors - Columns: band (predictor name), importance (Gini score) 2. CART_importance.csv - CART Gini importance scores for 15 spectral predictors - Columns: band (predictor name), importance (Gini score) 3. SVM_importance.csv - SVM permutation-based importance (accuracy drop method) - Columns: band (predictor name), accuracy_drop 4. RF_stability_5seeds.csv - RF classification stability across 5 random seed configurations - Columns: seed, OA (overall accuracy), Kappa, Luas_Ha (mangrove area) Study Area: Sumbawa Regency, West Nusa Tenggara, Indonesia Satellite Data: Sentinel-1 (SAR) + Sentinel-2 (Multispectral) Platform: Google Earth Engine Year: 2025