Author(s)

Zhisong Huang, Huizan Wang, Ren Zhang

Institute of Meteorology, PLA University of Science and Technology, Nanjing, China.

In Kriging interpolation, the types of variogram model are very finite, which make the variogram very difficult to describe the spatial distributional characteristics of true data. In order to overcome its shortage, an improved interpolation called Support Vector Machine-Kriging interpolation (SVM-Kriging) was proposed in this paper. The SVM-Kriging uses Least Square Support Vector Machine (LS-SVM) to fit the variogram, which needn’t select the basic variogram model and can directly get the optimal variogram of real interpolated field by using SVM to fit the variogram curve automatically. Based on GODAS data, by using the proposed SVM-Kriging and the general Kriging based on other traditional variogram models, the interpolation test was carried out and the interpolated results were analyzed contrastively. The test show that the variogram of SVM-Kriging can avoid the subjectivity of selecting the type of variogram models and the SVM-Kriging is better than the general Kriging based on other variogram model as a whole. Therefore, the SVM-Kriging is a good and adaptive interpolation method.

Least Square Support Vector Machine; Kriging Interpolation; Variogram; SVM-Kriging

Huang, Z. , Wang, H. and Zhang, R. (2012) An Improved Kriging Interpolation Technique Based on SVM and Its Recovery Experiment in Oceanic Missing Data. American Journal of Computational Mathematics, 2, 56-60. doi: 10.4236/ajcm.2012.21007.