engleski

Regression Models for Geospatial Big Data

There are different approaches that seek to model heterogeneity present in the spatial data. Spatial heterogeneity expresses different influence of independent variables on the dependent variable at the observations’ locations. These relationships can change abruptly between smaller areas within the total observed spatial area, while within these smaller areas they remain relatively homogeneous. They are important for data visualization and prediction modeling. To test different approaches of efficient modelling of spatial heterogeneity, the following models have been implemented and evaluated: classical multiple linear regression that models dependent variable as a single linear function of several independent variables (GLR), geographically weighted regression (GWR) model proposed in [1], geographically time weighted regression (GTWR) model described in [2], its time-only based variant (TWR), and our custom implementations of geographically clustered regression (GCR) and geographically time clustered regression (GTCR) that adds time dimension to enforce better predictions.

regression ; big data ; geospatial data

nije evidentirano