【作者】 丁睿；

【导师】 李夕兵；

【作者基本信息】 中南大学 ， 人口、资源与环境经济学， 2010， 硕士

【摘要】 人工神经网络是一门新兴的边缘科学,是模拟人脑智能结构和功能而开发出来的非线性信息处理系统,具有学习能力、并行性、容错性及易于硬件实现等基本特征,主要用于解决模式分类、函数逼近和数据压缩三大问题。近二十年来,神经网络在价格预测领域得到了广泛的应用。铀矿市场是一个高度复杂的非线性动态系统,其变化规律即有一定的自身趋势性,又受经济、政治等诸多因素的影响。传统的定量预测方法在研究铀矿价格预测时面临着许多困难。因神经网络具有自组织、自适应等特点,能自动从历史数据中发掘有关价格变化的知识,因而非常适用于解决铀矿价格预测问题。然而,迄今为止,这样的预测模型在国内外还未见报道。本文的主要工作包括以下几个方面：(1)简要分析了铀矿价格的影响因素,应用神经网络的原理,采用多输入、单隐层和单输出系统,建立了国际市场铀矿价格的神经网络预测模型,并对其未来五个月的变化进行了预测；(2)研究了BP神经网络在MATLAB中的设计与实现,以及如何在MATLAB中创建BP神经网络,如何对网络进行初始化、训练和模拟,介绍了本文经常用到的一些MATLAB函数,并采用MATLAB编程实现了所设计的BP网络；(3)对铀矿价格进行了实例预测研究,研究结果表明,预测结果的精度较高,证明了本文所采用的研究方法是实用而有效的,验证了本文所建立的基于BP神经网络的铀矿价格预测模型的有效性和适用性。更多还原

【Abstract】 The artificial neural networks is a newly developed interdisciplinary. It is a non-linear information processing system to imitate the structure and function of human brain.It possesses learning ability, parallelism, robustness and easiness for hardware implementation. It mainly applies to pattern classification, function approximation and date compression. Neural networks have developed rapidly in the past twenty years and have got a wide application in the fields of price prediction.Uranium price market is a highly complicate nonlinear system, whose variation does not have its own regulation, but also is influenced by many other factors, such as politics and economy. The tradition prediction techniques based on statistics face difficulties in uranium price research. Neural network have the virtue of self-organization and adaptability and can excavate the valuable information from historical data. So it is suitable to solve the problem in uranium price prediction.However, by now, researches did not established the relevant prediction models for uranium price at present. The paper mainly includes the following works:(1)The paper briefy introduces some basic knowledge on neural network and effected factors of uranium price. Using single hidden layer and multiple output system, a BP network model was built up to forecast the trend of uranium price in the next five months;(2)The paper briefy introduces how to realize BP neural network in MATLAB,how to create, initiate, train and simulate a network and some functions usually used in MATLAB.And a program is written to realize the BP network;(3)The paper does a experiment on the uranium price in the recent twenty years.,proves that the model and research method is effective.Compared with the prediction results, the results is briefly accepted, demonstrating that the forecasting model based on BP network has a good ability to forecast and generalize.更多还原