Mixed geographically and temporally weighted regression for spatio-temporal deformation modelling. Issue 385 (4th July 2022)

• Record Type: Journal Article
• Authors: Yang, Zhijia; Dai, Wujiao; Yu, Wenkun; Shi, Qiang; Santerre, Rock
• Is Part Of: Survey review. Volume 54:Issue 385(2022)
• Page Start: 290
• Page End: 300
• Publication Date: 2022-07-04
• Subjects: MGTWR -- Local linear estimation -- Local spatio-temporal non-stationarity -- Deformation modelling -- Spatio-temporal prediction; Surveying -- Periodicals; Great Britain -- Surveys -- Periodicals; 526.9
• DOI: 10.1080/00396265.2021.1935578
• Languages: English
• ISSNs: 0039-6265

Abstract

Abstract : When the regression coefficient of independent variable has both global stationarity and spatio-temporal non-stationarity properties, the deformation model based on the geographically and temporally weighted regression (GTWR) will no longer be applicable. In order to resolve this problem, we propose an improved method to establish the spatio-temporal deformation model using mixed geographically and temporally weighted regression (MGTWR). In this method, both the global regression coefficient and the variable regression coefficient are selected for regression coefficient hypothesis test, and the local linear two-step estimation method is used to fit the MGTWR model. A dam deformation modelling example shows that the MGTWR model improves the average prediction accuracy by 57.6% compared to the GTWR model when the regression coefficients have both global stationarity and spatio-temporal non-stationarity properties.