A Spatio-Temporal Series Simulation and Prediction Method of Geography Based on SVR-CA Model

To study simulation and prediction of geographical phenomena changes, the data sets of remote sensing (RS) formedasspatio-temporal series which is used as experimental data to obtain sample points. Construction of a newkernel function was proposed for RS characteristics and realized the multi-category of data judgment by using Least Squares Support Vector Regression (LSSVR) model in this study. And then, the spatio-temporal kernel function was converted to the nonlinear transformation rule of cellular automata (CA). The CA model based on RS data was designed to solve the boundary problem of cellular space. The cellular state transition algorithm was implemented. In application, the RS data sets of eco-environment vulnerability evaluation from 2002 to 2010 were used to simulate the evolution process of eco-environmental vulnerability, and predict the vulnerability change. Finally, experimental results show that the CA model based on LSSVRcan obtain better simulation and prediction results.