基于互联网评论的股票市场趋势预测

发布时间：2018-07-14 18:42

【摘要】：随着网络在国内普及率飞速增长，网络信息量呈几何级数增长，其传播的速度更是其它渠道难以匹敌的，成为人们最重要的信息来源之一。网络也成为金融领域信息重要的集散地，尤其是WEB2.0技术的发展，论坛、博客、聊天室等可以提供互动的技术不断涌现，使投资者可以参与到网络信息的创造、传播及获取的各个环节。论坛是最受欢迎的网络社区之一，众多的投资者在股票论坛中交流信息，分享经验以辅助投资决策，因此对其中信息的获取是了解投资者心理及行为的重要途径。相比国外上百年历史的成熟的金融市场，成立仅二十余年的中国金融市场还处于发展阶段，监管制度不完善，投机者居多。众多投资者通过各种途径获取信息进行交易，作为获取信息的重要方式之一，对股票论坛的研究具有重要意义。行为金融理论认为投资者的心理及行为能够影响股票市场的表现，基于这一理论，本文对国内的股票市场进行了研究。本文提出了自动剔除领域无关评论的方法，成功剔除了84%的股票市场无关评论，并保留了90%以上的股票市场相关信息。本文对比了语义分析方法、机器学习方法及N-Gram方法三种情感分析方法，支持向量机结合信息增益的方法能够获得良好的实验结果。通过单只股票价格影响因素分析，建立股票价格预测模型，能够比较准确地预测股票市场的价格。我们分析了股票价格影响因素，并建立回归模型对其进行预测。结果显示，滞后股票收盘价，情感指数，机构评分、滞后新闻数量能投对股票收盘价格进行解释。通过对通讯行业进行单因素方差分析，，情感指数能够影响收益率及波动率。通过对上证指数及情感指数进行领先滞后分析发现，投资者情绪与滞后综合指数相关，与领先个股收盘价相关。
[Abstract]:With the rapid growth of the popularization rate of the network in China, the amount of network information is increasing in geometric series, and the speed of its dissemination is even more difficult to compete with other channels, so it has become one of the most important sources of information for people. The network has also become an important distribution center for information in the financial field, especially the development of Web 2.0 technology, forums, blogs, chat rooms, and other technologies that can provide interaction, so that investors can participate in the creation of network information. All aspects of dissemination and acquisition. Forum is one of the most popular online communities. Many investors exchange information and share experiences in stock forums to assist in investment decisions. Therefore, the acquisition of information is an important way to understand the psychology and behavior of investors. Compared with the mature financial market with a history of more than one hundred years abroad, the Chinese financial market, which has only been established for more than 20 years, is still in the developing stage, the supervision system is not perfect, and the speculators are mostly. As one of the most important ways to obtain information, it is of great significance to the research of stock forum. Behavioral finance theory holds that investors' psychology and behavior can affect the performance of stock market. Based on this theory, this paper studies the domestic stock market. In this paper, a method of automatically eliminating domain-independent reviews is proposed, which successfully removes 84% of stock market independent reviews and retains more than 90% of stock market related information. In this paper, three affective analysis methods, semantic analysis method, machine learning method and N-Gram method, are compared. The support vector machine combined with information gain method can obtain good experimental results. Through the analysis of the influencing factors of a single stock price, a forecasting model of stock price is established, which can accurately predict the price of the stock market. We analyze the influencing factors of stock price and establish a regression model to predict it. The results show that the closing price of the stock can be explained by lagging stock closing price, affective index, agency score and lagging news quantity. By analyzing the single factor ANOVA, the affective index can affect the yield and volatility. Through the leading lag analysis of Shanghai stock index and emotion index, it is found that investor sentiment is related to lagging composite index and to closing price of leading stock.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.51;F224

【参考文献】