摘要：An Automated Valuation Model (AVM) that seeks to attain predictive accuracy must take into account both spatial and temporal effects in the real estate market. A model structure that contains neither explicit spatial nor temporal variables is calibrated by a method that recognizes these variations in is calibration architecture. The method is conceptually similar to Geographically Weighted Regression (GWR) except that it extends into the temporal domain. The methodology is explained and results provided illustrating spatio-temporal variations in value.
