【作者】 李霞；

【导师】 苏剑波； 环宇平；

【作者基本信息】 上海交通大学 ， 控制工程（专业学位）， 2013， 硕士

【摘要】 随着经济全球化的发展，企业面对的市场竞争愈演愈烈。要想征服市场，企业必须理解市场经济的规则，把握先机，对市场的未来作出准确的预测。而有效的市场预测是建立在认真研究各种市场因素的基础上的，因此企业必须做好前期调研和资料收集工作，在完备的数据系统基础上进行科学的市场预测，才能保证预测的准确性。只有保证了预测的准确性，企业才能制定准确的经营战略，最终赢得市场。本文首先介绍了几种传统的市场预测原理和预测方法，详细分析了各自的优缺点。随着计算机技术的发展，市场预测可以借助一定的软件来实现模拟分析，由于模拟分析具有廉价、迅速和可重复等优点，不受时间、地点等因素的制约，所以计算机仿真技术在预测过程中得到了广泛应用。预测与决策的一体化将是未来预测研究的重点。产品的销售业绩与市场需求是紧密联系，不可分割的，而市场需求的变化是复杂的，影响企业产品销售业绩的营销因素又是多样的、非线性的。在现有科技水平背景下，对于复杂的非线性系统来说，很难建立精确的数学模型，但是做出区间意义上的预测和逼近真实值的预测是可能的。因此，我们将目光投向在处理缺失值和非线性问题时有着明显的优越性的神经网络研究领域。本文详细介绍了神经网络的研究与发展，分析其结构模型、互联模式、学习算法和信息处理能力，详细分析了BP神经网络的结构和δ学习规则，概括介绍了面向MATLAB工具箱的神经网络设计。通过对江苏安华警用装备制造有限公司2000-2011年警用手铐销售情况进行统计，并以此数据为基础，通过对MATLAB神经网络进行学习、训练、仿真，对该公司2012年的警用手铐销售量进行了预测和仿真。企业一般会选择开发新产品来扩大产品的市场占有率，因此，我们也做了新产品市场前景预测的研究。研究结果表明，神经网络应用于企业销售预测和新产品市场前景预测具有较高的预测精度和良好的泛化能力，为销售预测提供了一种新的思路，为企业经营者科学决策提供有效的技术支持。更多还原

【Abstract】 With the development of economic globalization, market competition is intensified, in order to conquer the market, the enterprise must understand the rules of market economy, grasp opportunities, make accurate predictions about the future of the market. And effective market forecast is based on a careful study of various market factors, so the enterprise must do a good job in research and data collection,then make scientific market forecast. Only guarantee the accuracy of the forecast, the enterprise can make correct marketing strategy, ultimately win the market.This article first introduces several traditional market prediction theory and forecast method, and analyzes their advantages and disadvantages in detail. With the development of computer technology, the market forecast can use some software to implement the simulation analysis, due to the simulation analysis has the advantages of cheap, fast and repeatable, and is not affected by factors (such as time, place), so the computer simulation technology is widely applied in the prediction process. Integration of forecasting and decision-making will be the focus of future prediction research.Product sales was mainly determined by market demand. And Changes in market demand were complex. The influencing factors of enterprise product sales is diverse, non-linear. In the existing level of science and technology, it is difficult to establish a precise mathematical model of a complex nonlinear system, but we can make a prediction interval sense and close to the real value of the forecast.The neural network has obvious advantages in dealing with missing values and nonlinear problems. This paper described the research and development of neural networks to analyze the structural model, interconnect model, learning algorithms and information processing capacity,and analysed of the structure of BP neural network and δ learning rule, and introduced the MATLAB toolbox oriented neural network design.On the basis of police handcuffs’ sales of JiangSu ANHUA Police Equipment Manufacturing Co., Ltd. from2000to2011, by learning, training, simulation MATLAB neural network,we have predicted and simulated the company’s police handcuffs sales in2012. Enterprises tend to choose to develop new products to expand product market share, therefore, we also do the prediction research of New product market prospect. Simulation results show that the neural network to forecast business sales forecast has high accuracy and good generalization ability, it provided a new way of thinking for the sales forecast, also provide effective technical support for enterprise managers’ scientific decision.更多还原