CAJ | 학술논문

行为金融理论的研究表明,股票投资者在进行投资决策时,容易受到自身的因素如情绪与心理因素的影响。以行为金融理论为依据,作出基本假设：微博情绪信息反映的社会整体情绪倾向能够影响并预测股票市场整体价格走势的变化。实证过程包括抓取新浪微博数据并进行预处理,生成情绪倾向时间序列,通过格兰杰因果关系检验上证指数时间序列与情绪倾向时间序列间的相关关系,建立支持向量机模型预测股票市场价格的变化来验证假设的正确性。实验结果显示加入微博情绪信息的预测模型能够获得更高的准确率,进而证明了本文所作假设的正确性。
행위금융이론적연구표명,고표투자자재진행투자결책시,용역수도자신적인소여정서여심리인소적영향。이행위금융이론위의거,작출기본가설：미박정서신식반영적사회정체정서경향능구영향병예측고표시장정체개격주세적변화。실증과정포괄조취신랑미박수거병진행예처리,생성정서경향시간서렬,통과격란걸인과관계검험상증지수시간서렬여정서경향시간서렬간적상관관계,건립지지향량궤모형예측고표시장개격적변화래험증가설적정학성。실험결과현시가입미박정서신식적예측모형능구획득경고적준학솔,진이증명료본문소작가설적정학성。
Studies on stock price forecasting and investment decisions cover a very wide range of issues. One of them is the information efficiency of the market where the stocks are traded. The strong efficient-market hypothesis（ EMH） asserts that traded stock price reflects not only the economic value of stocks but also behavioral reaction of investors to the stock. Findings on behavioral finance theories show that investors are susceptible to emotional and psychological factors. The information is reported by not only some traditional media,such as newspapers,periodicals,radio,and television,but also some modern media,such as microblog.Hence,we propose a hypothesis that emotion tendency information shared on microblogs is helpful in forecasting stock price changes. Firstly,we collect massive microblog data using open API（ Application Program Interface） provided by Sina Microblog Platform. A total of 28,720,153 microblogs are collected in the time period from August 14,2009 to February 28,2012. Secondly,we classify microblog data according to its published date. We eliminate junk blogs,such as forwarded microblogs,microblogs with only link address in context,and other blogs that cannot reflect any emotions of the publisher. Thirdly,we analyze microblog emotion information using the semantic analysis tool ROST,and generate daily emotion tendency time series. We also collect the time series of Shanghai Composite Index in the same time period for comparison. The time series are standardized into Z-scores so that they can be compared with each other. Shanghai Composite Index shows correlation with some emotion time series. Fourthly,we use the Granger causality test to examine the correlation between differentiated Shanghai Composite Index and differentiated emotion tendency time series. Empirical evidence shows that there is a significant positive correlation between changes in Shanghai Composite Index and changes in the high passion tendency（ represented as DPC in this paper） series.Finally,we establish a nonlinear predicting model of the stock index by using support vector machine. Samples are collected from Shanghai Composite Index. Data in the trading days between July 1 and Nov 30 in 2011 are samples and those between December 1 and December 30 in 2011 are testing samples. Closing prices of Shanghai Composite Index in the lag 4 days are used to compose reference feature vectors. In addition,change of high passion tendency in the lag 4thday is added to form testing vectors. Empirical evidence shows that the predicting model with emotion information is more accurate.To conclude,the empirical evidence confirms our proposed hypotheses that the emotion information reflected in the microblog words improves accuracy in predicting changes in stock price.