论文部分内容阅读
为了提高害虫预报的准确率,将径向基小波网络首次引入农作物害虫预测预报领域,改进了径向基小波网络的学习算法,使之适合于害虫预测的应用:利用径向基小波函数族时、频两域支撑完全或部分覆盖被分析数据序列时、频两域支撑的原理来确定小波函数族尺度参数和平移参数取值;根据中心向量之间的欧式距离大小来初步筛选隐含层神经元。在实例分析中,本文利用1966-1995年山东省惠民县棉铃虫Helicoverpa armigera的监测数据建立了基于径向基小波网络的2代棉铃虫卵量峰值日期预测模型,利用1996-2000年的监测数据对模型进行了检验。检验结果表明:在5年的预测数据中,4年的预测数据偏差在3d以内,另外1年的预测数据偏差4d,预测效果令人满意。本文为害虫预测预报研究提供了一种可行的新方法。
In order to improve the accuracy of pest prediction, the radial basis wavelet network is first introduced into the field of crop pest forecasting, and the learning algorithm of radial basis wavelet network is improved, making it suitable for the application of pest prediction: , The two-frequency bistatic support fully or partially covers the data sequence under analysis and the principle of frequency bistatic support to determine the scale parameters and the translation parameter values of the wavelet functions. The preliminary screening of the hidden layer nerves according to the Euclidean distance between the center vectors yuan. In the case study, this paper used the monitoring data of Helicoverpa armigera from 1966-1995 in Huimin County, Shandong Province to establish the second-generation bollworm egg peak date prediction model based on radial basis wavelet network. By using the monitoring data of 1996-2002 The data tested the model. The test results show that in the 5-year forecast data, the deviation of the 4-year forecast data is within 3d and the forecast data deviation of the other year is 4d, the forecasting result is satisfactory. This article provides a feasible new method for pest forecasting research.