【作者】 贺清碧；

【导师】 胡久永；

【作者基本信息】 重庆交通学院 ， 管理科学与工程， 2004， 硕士

【摘要】 人工神经网络理论是最近发展起来的十分热门的交叉学科,有着非常广泛的应用前景。由于人工神经网络具有大规模并行信息处理,良好的自适应与自学习等许多特点,因此利用神经网络解决复杂非线性动态系统的预测问题就有了一条新的可能途径。基于神经网络预测方法的研究是当前预测研究中的一个重要课题,具有重要的理论意义和实际应用价值。本文在第一章介绍了人工神经网络理论的发展历史和应用领域,然后介绍了人工神经网络的特点及目前应用比较广泛的几种典型的神经网络。第二、三章着重研究了BP神经网络结构、算法原理,收敛性问题。第四章分析讨论了BP神经网络存在的问题。针对BP网络的缺陷,介绍了目前已有的一些有关BP网络改进的研究,同时,本文提出了一种改进的BP算法。在第五章本文将改进的BP算法应用于城市环境质量评价问题,建立了相应的BP网络模型(输入层神经单元个数6,隐层神经单元个数4,输出层神经单元个数1)。文中同时讨论了网络模型中输入数据的标准化处理问题,隐层神经单元个数的选取问题,以及网络参数中动量因子和学习率的选取等问题。并设计了基于BP神经网络的城市环境质量评价系统。该系统采用Visual FoxPro 6.0编制。实验结果与其它方法比较表明,该改进的BP算法收敛速度较快,具有较理想的预测精度,能满足实际应用的要求,是一种较为有效的预测方法,同时该BP网络预测系统也可用于相关的评价。更多还原

【Abstract】 Artificial Neural Network, a new but hot cross discipline, enjoys quite a great prospect of application. With its many characteristics such as large-scale parallel information processing, excellent self-adaptation and self-learning, the Network has been used in solving the complex nonlinear dynamic system prediction. The prediction study based on ANN is a important problem, enjoys significance in theory and reality.This paper has at chart one introduced the Network’ s history of development, its fields of application and then characteristics as well as the few typical neural networks used widely at present. At chart two and chart three, it has focused its research on the problems of back-propagation (BP) network structure, algorithms principle and convergence. At chart four, it has analytically discussed some of the problems that lie in BP Neural Network. In view of the flaws in BP Neural Network, the paper has introduced some improved BP algorithms and forwarded ones. At chart five, this paper applied it to the assessment of urban environmental quality, built a BP model, normalized the given data and decided the hide codes and parameters such as momentum and learn rate, and so on. At the same time, a assessment system of urban environmental quality based on ANN has designed with Visual FoxPro 6.0. The results of application show that the arithmetic is of high speed for convergence and of high accuracy for prediction and therefore meets with actual requirements of application and is an effective predictive method. This assessment system may be applied to the related assessment.更多还原