数据挖掘在股票预测中的应用教程.docx

天津理工大学 《数据挖掘技术》课 程 报 告 数据挖掘在股票预测中的应用 学院： 计算机与通信工程 专业： 计算机技术 姓名： 郑春园 学号: 153131314 数据挖掘在股票预测中的应用 摘要:作为海量数据的处理方法，数据挖掘从诞生起就和股票市场有着密不可分的联系。证券行业更是在信息化迅速发展的前提下成长起来的，很多业务都需要依赖对大量历史数据的分析，从中挖掘出有价值的信息。数据挖掘能够很好的处理这些非平稳、非正态、高噪声的数据。通过结合统计学的抽样、估计和假设检验；机器学习的搜索算法和建模技术；再加上最优化、信息论和信息检索等领域的思想，数据挖掘在处理海量数据和高频、多维数据上有着相当优势。最关键的是他能够对不断获得的新数据进行模型的动态更新，非常适合应用于新环境。 本文使用了时间序列方法人工神经网络模型对证券市场的价格进行建模和预测。应用时间序列方法预测时使用了简单一次移动平均预测法、一次指数平滑预测法、两次指数平滑预测法。时间序列方法预测股票走势最为简单，但是误差过大。应用人工神经网络时，采用了BP算法和遗传算法来训练网络权值，同时使用动量法和学习率自适应调整相结合的策略，人工神经网络模型虽然预测效果最好，但是计算量过大。 关键字：数据挖掘 股票预测 时间序列预测 人工神经网络 Abstract：As a huge amount of data processing methods, data mining and the stock marketis inextricably linked. The securities industry rapidly development because of theinformation technologys improvement. Many businesses in stock market rely ondigging out the large amount of historical data analysis to find the valuableinformation. Data mining can be very good to deal with these non-stationary,non-normal, high-noise data. By combining statistical sampling, estimation andhypothesis testing, machine learning search algorithms and modeling techniques, data mining have a considerable advantage in dealing with large amounts of data 、high-frequency, multi-dimensional data. The most important factor is that it cancontinue to embody the new data to dynamic update the data model which is verysuitable for the new environment. It is use the time．series methods．the artifieial neuralnetwork model in this papcr．AppIications of time-series forecasting methods have been used in a simple moving average forecast．a forecast of exponential smoothing,two exponential smoothing prediction method．time-series methods have the mostsimple in forc-easting trend of the stock,but have the largest error．Application ofArtificial Neural Network,BP algorithm and used genetic algorithms to train thenetwork weights，but also adopted the learning rate and m