【作者】 孔志周；

【导师】 方纪旋；

【作者基本信息】 湖南大学 ， 管理科学与工程， 2004， 硕士

【摘要】 销售预测在提高企业的经济效益及决策支持水平方面占有重要的地位。随着企业信息化水平的提高，企业销售数据的日益丰富，管理者对其中隐藏的销售预测信息的渴望日益强烈。用传统的方法来分析这些海量数据中的销售信息非常困难，已不能适应时代的要求。因此，如何找到更好的方法挖掘出销售数据中隐藏的销售预测信息是一个非常重要而值得研究的课题。 本文的研究工作源于上述背景。论文以数据挖掘技术为手段，小波神经网络为工具，对销售预测应用进行了研究，主要做了以下几方面工作：针对新形式下企业对销售预测的需求，提出了将数据挖掘技术应用于销售预测是必然的趋势，并在前人的基础上构建了适合销售预测的数据挖掘系统框架；针对销售预测中两个典型问题——销售量预测和潜在销售对象预测现有方法的缺陷，以小波神经网络作为数据挖掘的新方法，建立了适合这两个问题的小波神经网络预测模型；针对目前小波神经网络常用的BP学习算法易陷入局部极小的缺陷，采用遗传BP混合学习算法对小波神经网络进行学习；最后，为了检验本文方法的有效性和可用性，采用真实的销售数据对这两个典型问题进行了实证分析。 从结果及性能比较来看，采用了小波神经网络的预测方法效果是明显的，可以从海量的销售数据集中发现隐藏的销售量信息和潜在销售对象信息。当然，数据挖掘技术在销售预测中的应用还有许多问题值得探讨，本文的工作是一个尝试，还有待进一步深入。更多还原

【Abstract】 Sale prediction holds an important position in aspects of increasing enterprises’ economic profit and decision-making supporting level. With the improvement of enterprises’ informationalization level and the increasingly richness of sale data, the managers’ aspiration for hidden information on sale-prediction is increasingly strong. To analyze the sale information in the data of great capacity by traditional methods is very difficult and can’t suit the requirement of the era any more. Therefore, it’s a very important and worthy task to find a better method to mine the information on sale prediction hidden in sale data.The paper’s research is based on the above said background. The paper carried out research in application of sale prediction with the data mining technique as means and Wavelet Neural Network as tool. The following work has been done: 1) Aiming at enterprises’ requirement on sale prediction under new situation, we put forward that the application of data mining technique in sale prediction is an inevitable trend and construct a data mining system frame suitable for sale prediction based on predecessors. 2) Aiming at the limitation of existing method to predict sale volume and potential customer which are the two typical issues in sale prediction, we build wavelet neural network prediction model adapt to the two issues with the wavelet neural network as new method of data mining. 3) Aiming at the limitation of Back-propagation Algorithm commonly used in wavelet neural network at present, we study wavelet neural network by Genetic Back-propagation Algorithm. 4) We carried out positive analysis on the two typical issues by actual sale data in order to verify the validity and usability of the method proposed in the paper.From the result and the comparison of performances, the effect of prediction method by wavelet neural network is evident. The hidden information on sale volume and potential customer can be found in the set of sale data of great capability. Certainly, there’re still many issues worthy of discussion in the application of data mining in sale prediction. The work done in the paper is only an experiment awaiting further research.更多还原