基于网络舆情的SVM股票价格预测研究

发布时间：2018-02-22 19:47

本文关键词： 网络舆情支持向量机股票价格预测　出处：《南京信息工程大学》2014年硕士论文　论文类型：学位论文

【摘要】：自证券市场建立以来,作为高收益和高风险并存的股票,一直是众多投资者关注的对象。随着互联网络平台的快速发展,大数据时代到来,传统的股票技术指标数据已不能满足人们分析预测股票价格的需求。本文提出一种基于网络舆情和股票技术指标数据的支持向量机回归模型(NPO-SVM),该模型提高了股票价格的预测精度。模型首先抓取股吧、微博等股评信息,将这些股评观点用支持向量机算法分为看涨、看跌、看平三种股评情感倾向,计算看涨股评观点占看涨股评和看跌股评观点的比例作为网络舆情；然后对网路舆情以及与股票收盘价相关系数在0.6以上的股票技术数据作主成分分析,最后对保留的主成分运用支持向量机回归模型预测。并与基于股票技术指标数据的支持向量机回归模型(TI-SVM)以及基于经验模态分解的支持向量机回归模型(EMD-SVM)作对比,实证分析四只具有代表性的股票,得出NPO-SVM模型比TI-SVM模型、EMD-SVM模型具有更高的预测精度,可为股票投资者提供一种可靠的预测股票价格的方法。本文主要研究工作如下： (1)提出了一种将股评文本信息利用SVM机器学习,实现文本信息情感分类的新方法。该方法能够将海量(日均百万条)文本信息准确分类,测试分类准确率为85.4%。计算文本分类后的网络舆情值,得出网络舆情与股票收盘价之间的相关系数为0.7,说明网络舆情与收盘价之间的相关性较强。 (2)提出了一种基于网络舆情和股票技术指标的支持向量机回归模型,对股票收盘价预测。实证分析结果表明,NPO-SVM模型的最大相对误差为2.7%,平均绝对误差为0.092,平均相对误差为0.7%,趋势正确率为76.37%。与TI-SVM模型、EMD-SVM模型相比,NPO-SVM模型的预测精度明显提高。
[Abstract]:Since the establishment of the securities market, as a stock with high yield and high risk, it has always been the object of attention of many investors. With the rapid development of the Internet platform, the era of big data has come. The traditional stock technical index data can not meet the demand of people to analyze and forecast the stock price. This paper presents a support vector machine regression model based on network public opinion and stock technical index data. The model improves the precision of stock price prediction. These points of view are divided into bullish, bearish and leveling three kinds of stock review emotional tendency by using support vector machine algorithm, and the proportion of bullish and bearish opinion is calculated as network public opinion. Then the principal component analysis is made on the network public opinion and the stock technical data with a correlation coefficient of 0.6 or more with the closing price of the stock. Finally, support vector machine regression model is used to predict the retained principal components, and compared with the support vector machine regression model (TI-SVM) based on stock technical index data and the support vector machine regression model (EMD-SVM) based on empirical mode decomposition. Through the empirical analysis of four representative stocks, it is concluded that the NPO-SVM model has higher prediction accuracy than the TI-SVM model and can provide a reliable method for stock investors to predict the stock price. The main work of this paper is as follows:. This paper proposes a new method to classify the text information by using SVM machine learning. This method can classify the massive text information (millions of text information per day) accurately. The accuracy of test classification is 85.4. The correlation coefficient between network public opinion and stock closing price is 0.7, which shows that the correlation between network public opinion and closing price is strong. (2) A support vector machine regression model based on network public opinion and stock technical index is proposed. The results of empirical analysis show that the maximum relative error of NPO-SVM model is 2.7, the average absolute error is 0.092, the average relative error is 0.7, and the trend accuracy is 76.370.Compared with the TI-SVM model EMD-SVM model, the prediction accuracy of this model is obviously improved.
【学位授予单位】：南京信息工程大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F830.91;F224

【参考文献】