Spatial Prediction Models for Mining Spatial Data

摘要

The multivariate linear regression (MLS) model is a very good technique for non-spatial prediction. But spatial prediction needs to account for spatial information, which makes the MLS model inappropriate, for it assume that the learning samples are independently and identically distributed (i.i.d). Due to account for spatial information, the spatial auto-regression (SAR) model can be used for spatial prediction, but it is computationally very expensive. In this paper, we add spatial information into input variables by replacing each input variables with the weighted average of its neighbors and feed the new input variables to a MLS model to estimate model parameters, and then make spatial prediction, where MLS{sup}* stands for this model. Experimental results show that the MLS{sup}* model and the SAR model have almost identical effects on spatial prediction, while the MLS{sup}* model is computationally more efficient than the SAR model.