论文部分内容阅读
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.
According to the small samples and non-linear characteristics of pest occurrence data, a method of pest prediction based on least square support vector machine is proposed.Multiple linear regression analysis is used to select the influencing factors of pest occurrence, and then genetic algorithm is applied to least squares Support vector machine parameters optimization, and finally establish a complex non-linear relationship between the pest occurrence and impact factor model.Establish the model using one hundred larvae of the second generation corn borer test results showed that, compared with the multiple linear regression, BP The neural network model and least squares support vector machine (SVM) improve the prediction accuracy of the second generation corn borer, which is an effective prediction method for the change of pests.