摘要
在大数据环境下,为了提高农作物产量预测精确度和运算速度,提出首先基于Spark框架下处理海量数据方法,提高大数据处理速度。然后,运用上代精英位置组合策略实现个体增强优化DA算法(即EIDA算法),使DA算法摆脱收敛早熟困境,再利用EIDA很强的全局搜索能力帮助BP算法找出最佳初始化权值和阀值,避免其陷入局部极小值,提高BP算法精确度。实验结果表明:基于Spark框架下的EIDA-BP算法的农作物产量预测,不管是速度还是精确度都比其他类型的BP神经网络预测的高。
In the context of big data,in order to improve the accuracy and computing speed of crop yield prediction,the method of processing massive data based on Spark framework was first proposed to improve the speed of big data processing.Then,the last generation elite position combination strategy was used to realize the individual enhancement and optimization DA algorithm(EIDA algorithm),make DA algorithm get rid of the dilemma of premature convergence,the strong global search ability of EIDA was used to help BP algorithm find out the optimal initialization weight and threshold,avoid getting into local minimum and improve the accuracy of BP algorithm.Experimental results show that the speed and accuracy of the EIDA-BP algorithm based on Spark are better than those of other BP neural networks.
作者
唐立
李六杏
王启亮
王睿
方政
TANG Li;LI Liuxing;WANG Qiliang;WANG Rui;FANG Zheng(Department of Information Engineering,Anhui Institute of Economic Management,Hefei 230031,China;School of Information Management,Anhui Agricultural University,Hefei 230031,China)
出处
《邵阳学院学报（自然科学版）》
2020年第2期88-95,共8页
Journal of Shaoyang University：Natural Science Edition
基金
安徽省社会科学联合会课题(2018CX104)
安徽省高等学校自然科学重点项目(KJ2019A0965)。
