About Cross-Validation
Before you establish your final estimation, you may hope to know if your obtained hypothesis of the unknown points is reasonable or not. The Cross Validation can provide you the whole distribution of estimated values to help you make your final decision to your estimation selection.
Cross Validation estimates all points one in one. It selects a point first and abandons its value, and then estimates the value of the selected point (location) comparing with other values of other points. Through Cross Validation, you can compare the differences between estimated values and expected values and can also provide more data when you select an estimation model, such as Ordinary Kriging.
Cross Validation estimates the autocorrelation model with all the data. As the graph below, for example, there are 10 sample points in this area. Cross Validation has to locate a point (the red dot in the left graph) and abandons its observed value. Then it can estimate this red point by comparing with the other 9 blue points. The comparison between the estimated and measured values of the same location (the location of red point) is hence established. The second round runs in the same way. Cross Validation has to locate another point (the red dot in the right graph), abandons its observed value and then processes estimation and comparison. After all the points are re-estimated, part of the data may be assigned to different locations. They may need to be relocated through the autocorrelation analysis model.
©2017 Supergeo Technologies Inc. All rights reserved.