Extracting and Visualizing Location Effects with Machine Learning in Mass Valuation
Presentation Category
GIS Technologies
General Session Description
Quantifying the influence of location remains one of the most persistent challenges in mass appraisal. This presentation compares several data-driven methodologies for deriving location adjustments, focusing on the use of Gradient Boosting Models (GBMs) and traditional geospatial techniques. GBMs are leveraged to model sale prices and generate property-level “Location Factors” (LocFs) using SHAP (SHapley Additive exPlanations) values, which isolate and quantify the spatial component of value. These SHAP-based LocFs are compared with those derived from geostatistical and regression-based approaches, highlighting tradeoffs between interpretability, accuracy, and practical application. Visualization of LocFs reveals localized pricing patterns and supports improved market segmentation and equity analysis. The study demonstrates how integrating explainable machine learning with spatial analytics can modernize traditional valuation workflows and enhance transparency in location-based adjustments.
Start Date
3-31-2026 4:00 PM
End Date
3-31-2026 5:00 PM
Extracting and Visualizing Location Effects with Machine Learning in Mass Valuation
Quantifying the influence of location remains one of the most persistent challenges in mass appraisal. This presentation compares several data-driven methodologies for deriving location adjustments, focusing on the use of Gradient Boosting Models (GBMs) and traditional geospatial techniques. GBMs are leveraged to model sale prices and generate property-level “Location Factors” (LocFs) using SHAP (SHapley Additive exPlanations) values, which isolate and quantify the spatial component of value. These SHAP-based LocFs are compared with those derived from geostatistical and regression-based approaches, highlighting tradeoffs between interpretability, accuracy, and practical application. Visualization of LocFs reveals localized pricing patterns and supports improved market segmentation and equity analysis. The study demonstrates how integrating explainable machine learning with spatial analytics can modernize traditional valuation workflows and enhance transparency in location-based adjustments.