联系我们

论文查询结果	返回搜索

论文编号：	641
作者编号：	2120061939
上传时间：	2008/6/20 9:59:03
中文题目：	基于粗糙集理论的股票价格预测研
英文题目：	Stock Price Forecasting Resear
指导老师：	安利平 副教授
中文关键字：	粗糙集 证券市场 股票预测tl
英文关键字：	Rough Set Securities Business
中文摘要：	2007年8月10日，我国股票市场的总市值超过2006年国民生产总值。资本市场中的股票市场在中国的地位越来越重要。股市行情不但是经济运行的“晴雨表”，而且是国民经济的报警器；它不但可以帮助上市公司完成融资，而且能为机构投资者和个人投资者理财提供新渠道。由于股票的收益与风险往往正相关，所以人们一直孜孜以求，探索其内在规律，寻找有效的分析方法和工具。因此，对股票市场内在规律的研究和预测具有极其重要的理论意义和应用价值。 但是，股票市场是一个高度复杂的非线性动态系统，其变化不但受政治、经济、心理、气候等诸多因素的影响，而且自身的运行有一定的规律，传统的预测方法已不能完全满足这种需要，而新的数据挖掘的方法能在一定程度上避免传统方法的缺陷，在应用中逐步显示其优越性。 证券市场上，某只股票的走势看起来杂乱无章，但实际上有其内在的变化规律。市场有效性理论和周期理论、波浪理论等前人的经验性的理论成果也说明了股价的可预测性。而这正是数据挖掘方法应用在股票上，对投资回报率进行预测的基础所在。本文在介绍分析股票基本理论、概念和预测面临的问题的基础上；在回顾、比较各种使用数据挖掘方法预测股票价格的基础上，选择经典粗糙集方法，对某只股票进行分析、建模和预测。最后，还对不同投资期长度的结果进行了比较。 本文使用数据挖掘的CRISP-DM方法，运用规范的流程研究了股价预测问题。在应用经典粗糙集方法中，使用ROSETTA软件，综合使用专家经验法、布尔推理和等频区间法对属性进行离散化，对比遗传算法、动态约简算法、穷举法和约翰逊算法对属性进行约简和规则获取，最后使用标准投票法，评价所获规则集的预测效果。最后我们以江西铜业(600362)为例对所建的预测模型进行规则提取，并用所获规则预测投资于该股票的收益率，取得了较好的效果。
英文摘要：	In August, 10th 2007, the total value of the stock market had exceeded gross domestic product (GDP) of the year 2006 in China. Stock market’s position has become more and more important. The stock market is the weather glass and alarm of economy, it not only help the enterprises acquire the capital they need and also provide new Channels for institution investors and personal investors. For capital incomes and risks are positive correlated, people are diligent in an attempt to explore its rules and regulations, to find effective analysis methods and tools to forecast it. Therefore, analysis and forecasting stock fluctuating rules and regulations would have important theory meanings and application values. The stock market is a highly complicated nonlinear dynamic system, and the internal regulations have its own trend, which is affected by a number of elements such as politics, economy and psychology. The traditional methods of forecasting can't meet the need in these situations, while the data mining methods have show its charm and advantages in practices. Stock market trend has its own rules in despite of seemingly disorderly and unsystematic. Efficient Market Theory, Wave Theory and Cycle Theory indicate that stock price is predictable, which is the base of appling the data mining. This thesis firstly introduces the basic conceptions and the facing problems in stock prediction. Then, we use classic rough set method to analyze, model and forecast ROI of the investment after retrospect and comparison of the various historical methods.Finally, we compare the different result of the different investment duration and different investment strategies. This thesis takes the standard method of the CRISP-DM to analysis, model and forecast stock price, found some rules and regulations, helping us in practices. With the help of the software ROSETTA, we apply the classic rough set method to stock data, discretizaiton the attributes with expert’s experience method, Boolean reasioning algorithm, Equal frequence binning algorithm; reduce attributes and generate rules with Genetic algorithm, Dynamic Reduction algorithm, Exhaustive calculation and Johnson's algorithm. In the end, we evaluated the effect of the generated rules with standard voting method. Finally, we apply the above models with the stock 600362. With training set and testing set, we obtain preferable results.
查看全文：	预览  下载(下载需要进行登录)