云计算平台支持下的BP神经网络在洪灾损失评估中的应用研究

发布时间：2018-05-28 09:17

本文选题：洪灾损失评估 + Hadoop　；参考：《江西理工大学》2017年硕士论文

【摘要】：在我国洪灾属于最为严重的自然灾害之一,其频率高、影响范围广及经济损失大等特征已经严重制约了我国国民经济的发展,因此对洪灾经济损失进行科学有效的估算是必要的。但是近年来,由于人类的活动增多、洪灾损失评估的数据种类和数剧量不断增加,导致传统BP神经网络技术在洪灾损失评估应用中可能出现耗时过长、训练困难等问题。考虑云计算平台在处理大量数据方面问题的优越性与实用性,而现有洪灾损失评估相关研究还未在云计算平台下进行应用。因此,开展云计算平台支持下的BP神经网络在洪灾损失评估中的应用研究具有现实意义。本文选取江西省鄱阳湖区范围内的某县作为研究区域,主要研究内容如下:首先,阐述了洪灾损失评估及相关技术的国内外的研究现状,论述本文所运用的关键技术:Hadoop分布式计算框架及BP神经网络技术,可为本文洪灾损失评估应用研究提供理论基础。其次,运用数理统计的方法对原始数据进行收集与整理,结合洪灾损失理论选择能够反映洪灾损失情况的洪灾影响因子,并根据洪灾影响因子划分得到进行计算的样本数据与测试数据;然后,在BP神经网络算法基本结构的基础上,将其拆分成两大部分:首先是网络学习部分,其次是权值调整部分。根据算法的拆分,将其分别在Map函数与Reduce函数中实现,得到云计算平台支持下的Mapreduce-bp算法;最后,根据Mapreduce-bp算法,建立云计算平台支持下的Mapreduce-bp神经网络洪灾损失评估模型,利用该模型对本文研究区域2013年的洪灾经济损失进行应用,并得出最终估算结果。本文的研究结果表明,云计算平台支持下的Mapreduce-bp洪灾损失评估模型能准确、快速的对洪灾经济损失值进行估算,因此该模型在大数据量的情况下能为高效的进行洪灾损失评估工作提供新的解决思路。
[Abstract]:Flood disaster is one of the most serious natural disasters in China. Its high frequency, wide range of influence and large economic losses have seriously restricted the development of our national economy. Therefore, it is necessary to estimate flood economic losses scientifically and effectively. However, in recent years, due to the increase of human activities, the variety and amount of flood damage assessment data is increasing, which may lead to the problems of time-consuming and difficult training in the application of traditional BP neural network technology in flood damage assessment. Considering the advantages and practicability of cloud computing platform in dealing with a large number of data problems, the existing flood loss assessment research has not been applied in cloud computing platform. Therefore, it is of practical significance to research the application of BP neural network supported by cloud computing platform in flood damage assessment. In this paper, a county in Poyang Lake region of Jiangxi Province is selected as the research area. The main research contents are as follows: firstly, the current research situation of flood disaster loss assessment and related technologies at home and abroad is expounded. This paper discusses the key technology used in this paper: the distributed computing framework of: Hadoop and BP neural network, which can provide a theoretical basis for the application of flood damage assessment in this paper. Secondly, using the method of mathematical statistics to collect and collate the original data, combined with the theory of flood loss to select the flood impact factors which can reflect the situation of flood losses. Then, based on the basic structure of BP neural network algorithm, it is divided into two parts: first, the network learning part. Second is the weight adjustment part. According to the split algorithm, it is implemented in Map function and Reduce function respectively, and the Mapreduce-bp algorithm supported by cloud computing platform is obtained. Finally, according to the Mapreduce-bp algorithm, the Mapreduce-bp neural network flood damage assessment model supported by cloud computing platform is established. The model is used to study the regional flood economic losses in 2013, and the final estimated results are obtained. The results of this paper show that the Mapreduce-bp flood loss assessment model supported by cloud computing platform can estimate the economic loss of flood accurately and quickly. Therefore, this model can provide a new solution for flood damage assessment in the case of large amount of data.
【学位授予单位】：江西理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09;X43;TV87

【参考文献】