当前位置: 首页 > 文章 > 基于GRA/BPNN的农作物害虫发生量预测模型 农机化研究 2013 (6) 19-24
Position: Home > Articles > Model for Predicting Occurrence of Crop Pests Based on BP Neural Network and Principal Components Analysis and Its Application Journal of Agricultural Mechanization Research 2013 (6) 19-24

基于GRA/BPNN的农作物害虫发生量预测模型

作  者:
彭琳;杨林楠
单  位:
云南农业大学云南省高校农业信息技术重点实验室;上海大学计算机工程与科学学院
关键词:
农作物;害虫预测模型;灰色关联度分析;BP人工神经网络
摘  要:
针对农作物害虫灾害发生的差异性、突发性、随机性、多样性和不均匀性等特点,将人工神经网络、灰色关联度分析与主成成分析相结合,提出一个新的农作物害虫发生预测网络模型。首先,针对影响农作物害虫发生影响因子较多的问题,模型通过主成分分析方法将影响因子进行简化处理;同时,为了实验数据的相关性,采用了灰色关联度分析,排除实验与统计等方面的误差;最后,利用BP人工神经网络构建了农作物害虫发生预测模型,并以斑潜蝇为例,进行了试报检验。检验结果表明,模型应用于农作物害虫灾害发生预测具有较高的预测精度和良好的泛化能力。
译  名:
Model for Predicting Occurrence of Crop Pests Based on BP Neural Network and Principal Components Analysis and Its Application
作  者:
Peng Lin1,2,Yang Linnan2(1.College of Computer Engineering and Science,Shanghai University,Shanghai 200072,China;2.University Key Laboratory of Agricultural Information Technology in Yunnan,Kunming 650201,China)
关键词:
crop;prediction model of pest;GRA;BPNN
摘  要:
According to the characteristics on occurrence of insect pest disasters such as no homogeneity,difference,diversity,abruptness,randomness,etc.,this paper,in combination of the BP Neural Networks(BPNN),Grey Relational Analysis(GRA) and Clustering Analysis Method,put forward a new prediction network model for occurrence of crop pests.First,according to the problem that there are more influence factors which influence the occurrence of crop pest,the model use the GRA to simplify the influence factor;meanwhile,in order to realize the relativity of data,it use the Clustering Analysis Method,eliminate the error about experiments,statistic and so on;Finally,it uses the BPNN to build the prediction model for occurrence of crop pests,and use the example of Liriomya huidobresis Blanchard,do the test report and examination.The result of the examination shows the model has higher prediction accuracy and better generalization ability in the prediction of the occurrence of crop pest.

相似文章

计量
文章访问数: 6
HTML全文浏览量: 0
PDF下载量: 0

所属期刊

推荐期刊