1{2 "record_id": "b922ab81201881b47779769df9e76de174e13df7",3 "eligible": true,4 "short_ref": "Oliveira, Reyes-Bueno, Gonzalez & da Silva (2024/2025)",5 "country": "Brazil",6 "city": "Fortaleza",7 "property_type": "apartments",8 "price_type": "listed prices and actual sales (log price)",9 "n_obs": 43585,10 "sample_period": "2017 and 2021",11 "data_source": "Real Estate Market Observatory, Local Finance Secretary of Fortaleza",12 "methods_compared": [13 "MRA (OLS)",14 "spatial error regression",15 "MARS",16 "M5P",17 "random forest",18 "quantile random forest",19 "GBM",20 "LightGBM",21 "CatBoost",22 "XGBoost",23 "deep learning"24 ],25 "best_method": "XGBoost",26 "metric": "MAPE",27 "best_value": 8.96,28 "hedonic_value": 13.9,29 "test_set": "holdout",30 "features_ai": "none",31 "incremental_ai_gain": null,32 "incremental_ai_metric": "",33 "marginal_effects": "coefficients",34 "notes": "Appendix Table B, stratified random split 32,688 train / 10,897 test: MAPE 8.96% XGBoost, 9.05% RF, 9.25% GBM, 13.90% MRA, 11.02% SER; R2 0.96 for RF/GBM/LightGBM/XGBoost vs 0.88 MRA and 0.94 SER; RMSE in BRL (XGBoost 116,633.77; GBM lowest 113,640.60; MRA 197,400.28); COD 8.98% for RF and XGBoost. XGBoost is the only model with no residual spatial autocorrelation (Moran's I). Coefficients reported for MRA/SER only; ML models are not explained (authors defer interpretation to another study).",35 "confidence": "high"36}