Forecasting Pest Occurrence And Recognizing Insect Using Support Vector Machine

Combined support vector regression (SVR) and controlled autoregressive(CAR), a non-linear multidimensional time series approach named SVR-CAR thatbases on structural risk minimization and has high accurate prediction was constructed,which showed the dynamic characteristics of sample set as well as the effect ofenvironmental factors. And it is used to forecast the diseased panicle rate of wheatscab and the damage degree of the 2nd generation corn borer for five years. The resultsindicated that: the mean squared error (MSE) is 205.2 and the mean absolutepercentage error (MAPE) is 18 at the diseased panicle rate of wheat scab; the MSE is30.2 and the MAPE is 27 at the damage degree of the 2nd generation corn borer. TheSVR-CAR model is higher than the ones of all reference models and it has the wildapplications in many fields such as agricultural science, ecology and economics.Based on 11 math-morphological features (MMFs) (such as from area, perimeter,eccentricity, roundness etc.) extracted from the images of 34 species of insects of theHemiptera, Lepidoptera and Coleoptera, the application of suppose vectorclassification (SVC) was evaluated in insect taxonomy at different levers of order,family class and class. These 11 MMFs was optimized by continually pan out themost valueless one in order to increase precision. The results indicate that it is feasibleto identify insects by SVM, and that MMFs-optimizing is beneficial to the precision.