最小二乘支持向量机在害虫预测中的应用

Application of least square support vector machines in pest forecast

针对害虫发生量数据的小样本、非线性特点,提出一种最小二乘支持向量机的害虫预测方法.首先采用多元线性回归分析法选择害虫发生量的影响因子,然后通过遗传算法对最小二乘支持向量机参数进行优化,最后建立害虫发生量与影响因子之间复杂的非线性关系模型.采用二代玉米螟百株幼虫虫量对模型性能进行检验,结果表明,相对于多元线性回归、BP神经网络模型,最小二乘支持向量机提高了二代玉米螟发虫量的预测精度,是一种有效的害虫变化预测方法.

In view of pest occurrence’s small sample data,nonlinear characteristic,a pest forecast method was proposed based on least squares support vector machines.Firstly,the influence factors of pest occurrence area were selected by multiple regression analysis method,and then the parameters of least square support vector machines were optimized by genetic algorithm,lastly build the complex nonlinear model was built between pest occurrence and influence factors.The proposed model was tested by the second-generation Corn Borer’s occurrence,the results show that the proposed model improve the forecasting accuracy of the second-generation Corn Borer’s occurrence compared with the multiple linear regression and BP neural network;the proposed model is an effective forecasting method for pest occurrence.