Abstract

To improve the accuracy of forecasting the peak period of Dendrolimus punctatus, fuzzy comprehensive evaluation method, contingency table analysis and BP neural network were used to predict peak period of D. punctatus for the first generation egg from 1983 to 2016 in Qianshan County. The forecasted values for the three methods were in 2017 and 2018 were consistent with the actual results. The historical coincidence rate for fuzzy comprehensive evaluation method from 1983 to 2016 was 93.94% and for contingency table analysis was 100%. When taking into consideration the standard error of BP neural network was 4 d, the historical coincidence rate was 100%. Among three methods, The screening predictors closely related to forecasted quantity and the classifcation standard were the key for improving the accuracy of forecast.

摘要

为提高马尾松毛虫高峰期预测的准确性，采用模糊综合评价法、列联表分析和BP神经网络对马尾松毛虫高峰期进行预测。1983年至2016年潜山县第一代蛋。三种方法2017年和2018年的预测值与实际结果一致。1983-2016年模糊综合评价法的历史符合率为93.94%，列联表分析的历史符合率为100%。考虑到BP神经网络的标准误差为4 d，历史符合率为100%。在三种方法中，与预测量密切相关的预测因子筛选和分类标准是提高预测精度的关键。