Research On Nonlinear Forecasting Approach Of Crops Diseases

The crop disease is one of the most important disasters in Chinese agriculture. Atpresent, linear methods are the common methods to predict crop diseases. But theagricultural system is a very complicated nonlinear system. The forecast precisions of themodels are not very high by using general linear methods. It seriously affects the yield ofcrops. Meanwhile, people pay more and more attention to the quality and safety of foods.Currently, the way of control crop diseases is still using pesticide. If we can predict theoccurrence of crop diseases relatively, the amount of pesticides can be under the control.Then we can improve the crop quality. In recent years, with the continuous development ofnonlinear science, more and more nonlinear methods are invented by scientists. Nonlinearmethods can better reveal the nature and law about the nonlinear system of agriculture.And predictive models based on nonlinear methods have very important significance onimproving the prediction of crop diseases.This paper takes the rice blast of Jixi County as the research object. First, analyze thecauses of the rice blast. Because the occurrence of the rice blast has many nonlinearcharacteristics, such as uncertainty, multiple input complex, and so on. This paper chosetwo kinds of nonlinear methods to analyze and forecast the occurrence of the rice blast.The first method is the partial least squares regression. According to the rice blast’s maincauses, combined with the weather information, I choose temperature and precipitation asthe main meteorological factors to regress the model. The partial least squares regressionmethod has effectively solved the multicollinearity among variables, as well as the numberof sample points less than the variables. At last，I use MATLAB to program and realize thealgorithm. The precision of the model is basically satisfactory and it can effectively carryout the rice blast prediction accuracy.Secondly, the paper uses the chaos theory to analyze the rice blast’s time series of JixiCounty by the MATLAB chaotic time series box. Chaos theory is based on the phase spacereconstruction. First of all, find the best time delay and the best embedding dimension toreconstruct the phase space. And then distinguish whether the time series is chaotic. Lastly,build the prediction model.I use the auto-correlation function method and the mutualinformation method to calculate the delay time of phase space reconstruction, use Cao’smethod to obtain the embedding dimension. To calculate the largest Lyapunov exponent byusing the method of small data, which is a quantitative method to judge chaos. Finally, build the chaos-RBF neural network. The results showed that the predictions about smalltime series are effectively only on a certain steps.These two nonlinear methods, either partial least squares regression method or chaostheory, are all at the initial stage on our country’s crop diseases’ research. There are stillmany problems should be further improved. Through this paper’s models, we can see thatnonlinear methods in crop disease forecasting are feasible. It has certain significance on thecrop diseases’ forecasting development.