{
 "record_id": "ecb8bacfa12ff5fad71503a1b459329f9f420bc4",
 "eligible": false,
 "exclusion_reason": "Two-stage mass-appraisal model: k-means clustering into submarkets (on area, age, bedrooms, bathrooms, type) followed by ordinary kriging of prices within each submarket. No hedonic regression is estimated; the paper reports only cluster descriptive statistics (Table 2), variogram models and LOOCV fit statistics (R-squared, RMSE, MAPE; Tables 3-8) and kriging maps. No coefficient on any property characteristic, no SE. Prediction-only study; keep qualitatively (AI/geostatistical AVM evidence, Medellin, Colombia, 263 appraisals).",
 "study": {
  "title": "A House Price Modeling Based on Clustering and Kriging: The Medellín Case",
  "year": 2022,
  "doi": "10.52324/001c.66198",
  "country": "Colombia",
  "region_city": "Medellín",
  "sample_start": 0,
  "sample_end": 0,
  "property_type": "mixed",
  "price_type": "appraisal",
  "data_source": "Valuations (appraisals) of second-hand housing units, Medellín Property Price Register; 263 observations (54+65+144 in Case B)",
  "document_type": "journal",
  "estimators": ["other"],
  "spatial_method": "ordinary kriging (Matern, spherical, circular, Gaussian, exponential variograms) within k-means submarkets",
  "time_fe": false,
  "location_fe": false,
  "se_type": "unknown",
  "n_obs_total": 263,
  "ai_methods": ["k-means clustering", "kriging"],
  "notes": "Sample period not stated in extracted pages (Quality of Life Survey 2018 referenced). Prices in COP. Best configuration: 3 submarkets by area and age; LOOCV R2 -0.13/0.22/0.58, MAPE 26.8/42.3/32.9 % (Table 8, p.15)."
 },
 "models": [],
 "coefficients": []
}