Study Of Drought Prediction Based On Support Vector Machine

The drought is one of the main natural disasters. It is also one of the hot issues of the atmosphere of scientific research. This paper used the method of Support Vector Machine, and studied for based on climate factor of soil humidity mixed forecasting model and based on time series of Drought Prediction model. The results of the study will benefit to the widespread drought monitoring for government and relevant departments to provide the scientific basis of Drought Relief.The main conclusions are as follows:(1) This paper designed a model of soil relative humidity mixed forecasting which based on Support Vector Machine (SVM). and conducted the simulation experiments. And then according to the traditional support vector machine’s weaknesses on the method of choosing parameters, grid partition method. genetic algorithm and particle swarm algorithm are used to optimize the parameters of the model. The simulation results show that, the introduction of the SVM method provides a new tool for predicting soil moisture, in addition, the method which directly based on crop water signal is unable to achieve, and for the nonlinear problem solving provides a new idea.(2) This paper studied the time series prediction model which based on the SVM, established model after optimizing parameters, and conducted the simulation prediction with the data of one station’s average temperature and precipitation. And then based on the forecasting results calculated out the drought index, and then predicted drought level. The forecast results is accurate, shows that the drought forecast method is effective and practical.