分类号：TP391密级： 公开U D C ：单位代码： 10424高校教师在职攻读硕士学位论文基于小波网络的数据挖掘技术及在股市预测中的应用段申请学位级别：硕士学位文秀专业名称：计算机软件与理论指导教师姓名： 路 燕职称： 副 教 授山 东 科 技 大 学二零零六年十月论文题目：基于小波网络的数据挖掘技术及在股市预测中的应用作者姓名： 段 文 秀入学时间：2005 年 3 月专业名称：计算机软件与理论研究方向：数据库系统指导教师：路 燕职称：副 教 授论文提交日期：2006 年 10 月论文答辩日期：2006 年 12 月授予学位日期：THE APPLICATION OF DATA MINING TECHNIQUEBASED ON WAVELET NEURAL NETWORK IN STOCKMARKET FORECASTINGA Dissertation submitted in fulfillment of the requirements of the degree ofMASTER OF ENGINEERINGfromShandong University of Science and TechnologybyDuan WenxiuSupervisor: Associate Professor Lu YanCollege of Information Science and EngineeringOctober 2006声明本人呈交给山东科技大学的这篇硕士学位论文，除了所列参考文献和世所公认的文献外，全部是本人在导师指导下的研究成果。该论文资料尚没有呈交于其它任何学术机关作鉴定。研究生签名：日期：AFFIRMATIONI declare that this dissertation, submitted in fulfillment of the requirementsfor the award of Master of engineering in Shandong University of Science andTechnology, is wholly my own work unless referenced of acknowledge. Thedocument has not been submitted for qualification at any other academicinstitute.Signature:Date:山东科技大学硕士学位论文摘要摘要数据挖掘是一项较新的数据库技术，它基于由日常积累的大量数据所构成的数据库，从中发现潜在的、有价值的信息称为知识，用于支持决策。数据挖掘技术可用来进行关联分析、分类、聚类、预测、孤立点挖掘等。股票市场在我国产生以来不断成长，逐步成为证券业乃至整个金融业必不可少的组成部分，成为投资者、管理者和经济学者共同关注的热点。因而对股票市场走势的分析和预测具有重大的理论意义和可观的应用价值,但传统的线性统计预测模型的预测效果并不理想。随着非线性理论和人工智能技术的发展，小波分析和小波网络等成为金融市场强有力的分析和预测工具。本文致力于研究将小波网络引入数据挖掘技术以便对股票市场进行预测。在研究了股票分析技术、数据挖掘技术、小波网络的构造和学习算法的基础上，给出一种新的小波网络构造学习算法，建立应用于股市预测的非线性组合预测模型。主要工作如下：1、通过对影响股市价格波动的因素、股票现行预测方法的分析，论证了通过小波网络来对股市行情这种非线性的动力系统进行预测的必要性。2、通过对数据挖掘的相关背景知识进行研究，论证了数据挖掘技术是非常有应用前景的一种数据库技术，用此技术可以来对历史数据进行分析，以便预测数据的未来趋势。3、通过对小波网络的背景知识学习，给出一种新的小波网络的构造学习算法。通过实验，验证了此算法的有效性，并应用基于此算法的小波网络建立非线性组合预测模型，给出预测模型的原理和评价标准，选取训练样本对网络进行训练，得到合适的网络结构和组合函数后，对股市进行预测。关键词：数据挖掘，小波网络，股市预测，学习算法，非线性组合预测山东科技大学硕士学位论文摘要ABSTRACTData Mining is a new database technique which aims at discoveringpotential and valuable pattern that is called as knowledge from database.The knowledge is widely discovered can be used for decision-making. DataMing techniques can be used to association analysis, classification,clustering, forecasting and outlier mining, etc.Stock market became the most important and absolute necessary part ofsecurities trade and financial market after the establishment in China, andit is a hotspot that investors, administrators and economist pay attentionto . So the analysis and forecasting of stock market have not onlysignification of theory but also merit of application. But the predictionresults of traditional prediction technology are unsatisfied. As thedevelopmentofnonlineartheoryandartificialintelligence,waveletanalysis and wavelet network become cogent tools for money market analysisand forecasting.The paper lead wavelet neural network in Data Mining techniques ,anduse it to forecast stock market. Then put forwards the self-constructingalgorithm based on the research of the stock market analysis technology,Data Mining and construction of wavelet network and self-study algorithm;constructs the non-linear combination model used for the prediction ofstock market. The main work is:1. The necessity of forecasting stock market-the non-linear dynamicsystem though wavelet neural network is obtained through the analysis tothe factor which affects the stock market price and to the present forecastmethod analysis of stock market.2. It is obtained that Data Mining is a very useful database techniqueand the technique can be used for analysis old data and forecast the trendof the data after studying the background knowledge of Data Mining.3. The paper put forward a new study and constructing algorithm of thewavelet network through studying the background knowledge of wavelet neuralnetwork. The algorithm is proved effective by experiment, and use it onwaveletneuralnetworkwhichisbasedonthealgorithmofself-山东科技大学硕士学位论文摘要constructing to establish nonlinear combination forecasting model ,providethe theory and study algorithm and evaluation standard of