1{2 "record_id": "ee4a73efc5ad67b4400db23b09eeacbeb4812f86",3 "eligible": false,4 "exclusion_reason": "No hedonic coefficients printed: the paper reports only mean absolute SHAP values and permutation feature importances (Tables 3-4, temporal figures) and predictive-performance metrics (Appendix Table A1) for a log-linear regression and a random forest; the linear regression coefficients themselves are not reported (Appendix A2 gives VIFs only). Feature importances are not marginal effects and cannot enter the coefficient synthesis (keep qualitatively for the AI/AVM section).",5 "study": {6 "title": "Explainable AI-based mass appraisal: Insights from machine learning applications in Korea's residential property market",7 "year": 2025,8 "doi": "10.3846/ijspm.2025.25137",9 "country": "South Korea",10 "region_city": "Seoul, Gangnam District",11 "sample_start": 2006,12 "sample_end": 2017,13 "property_type": "condo_apartment",14 "price_type": "transaction",15 "data_source": "Apartment transactions in Gangnam District, Seoul, 2006-2017 with complex-level, locational and macro variables",16 "document_type": "journal",17 "estimators": [18 "OLS",19 "ML_marginal"20 ],21 "spatial_method": "none (Dong dummies)",22 "time_fe": true,23 "location_fe": true,24 "se_type": "unknown",25 "n_obs_total": null,26 "ai_methods": [27 "random_forest",28 "SHAP",29 "permutation_feature_importance"30 ],31 "notes": "AVM/XAI study: linear regression R2 0.892 (log price) vs random forest R2 0.970 (5-fold CV). Variables include area, construction year, floor level, parking, heating system, brand, complex scale, distances to subway/park/school/university/hospital/museum, and macro series. Only importance rankings are printed."32 },33 "models": [],34 "coefficients": []35}