【作者】 陈亮；

【导师】 戴伟辉；

【作者基本信息】 复旦大学 ， 软件工程（专业学位）， 2014， 硕士

【摘要】 随着现代金融市场与信息网络技术的不断发展,证券市场的交易手段和方式正在不断发生新的变化。股票市场作为影响最为广泛的证券市场,已经显现出大量的高频交易特征。上述发展新趋势引起了证券市场投资者、管理者及相关研究学者的高度关注,已成为证券市场研究领域的重要新热点。股票市场的高频交易在很大程度上要受到市场情绪效应、投资者非理性行为的影响和各类事件冲击所产生的噪声干扰,对传统的金融交易分析技术提出了新的挑战。近年来,很多研究学者发现互联网微博舆情所反映的市场情绪倾向及上述情绪的传播与股票高频交易之间存在着很强的相关性,其内在机制及微博舆情的情绪分析方法是股票高频交易分析技术中尚须深入探索和研究的关键问题。本文选题来源于复旦大学金融研究中心高端学术研究交叉学科课题(No.2012FDFRCGD02)。本文试图从行为金融学和情感计算角度构建基于微博舆情的股票高频交易分析技术新方法。本文首先从股票高频交易的影响因素入手,通过行为金融学分析并结合相关学者的实证结论,阐述了中国股市在较大程度上受消息面和投资者情绪影响的基本特征。然后,进一步对各类事件消息所引发的舆情进行了分类,对上述舆情的产生、情绪传播及由此所导致的交易行为这一过程和系统性机制做了深入剖析。在此基础上,论文运用情感计算技术对微博舆情的情绪分析方法做了研究。本文针对微博文本数据处理问题,基于自然语言处理思想,提出对表情、符号进行情感倾向分类,增加证券情感词典等数据处理的探索性尝试。本文针对微博舆情情感计算问题,基于SVM机器学习方法,提出微博舆情情感计算分析框架。本文针对高频交易分析技术的应用问题,立足于传统金融计量模型,加入情感计算变量,结合高频交易数据进行时序性分析,提出以微博情感计算结果为依据的高频交易分析模型。针对此模型作了要素和框架分析,并进行实证检验。阐述了软件系统框架设计、模型的数据流程及运行方式并通过软件实现。本文的研究成果为揭示股票高频交易的市场情绪传导及微博舆情的信息处理方法和相关分析技术提供了新的参考。更多还原

【Abstract】 As an on-going development of the modern financial market and information network technology, the means and ways of trading in the securities market are constantly making new changes. The stock market, as the most influential securities market, has shown a lot of characteristics of high frequency trading (HFT). This new trend of development has received a high level of attention from the investors and managers of the securities market as well as the research scholars of the related field. It has become an important new hot topic of the securities market research.Market sentiment, irrational behavior of the investors and noise generated by various events can all influence HFT greatly. This produces a new challenge to traditional financial transactions. In recent years, many researchers find out that there is a strong correlation between the tendency and the spread of market sentiment reflected by public opinion on Weibo and HFT. The internal mechanism and the method of analyzing the public opinion on Weibo are the key point to be explored and researched thoroughly in technical analysis of HFT.This paper selects topic from High Level Academic Research Program of Financial Research Center, Fudan University (No.2012FDFRCGD02). This article tries to construction a new method to analyze HFT based on the public opinion on Weibo, from the point of behavioral finance and affective computing. Starting from the influence factors of HFT, this article first elaborates the essential feature of Chinese stock market, which to a larger extent affected by the news and investor sentiment, based on the analysis of the behavioral finance and the empirical conclusions of related scholars. Then, this paper classifies public opinion caused by all kinds of event messages, and analyzes in depth the systematic mechanism and the process of how the trade behavior is impacted by the generation of the above mentioned public opinion and the spread of sentiment. Base on all of above, this paper researches the method of the sentiment analysis with the public opinion on Weibo.For the problem of text processing Weibo data, we propose an exploratory data analysis method using securities sentiment word dictionary based on natural language processing (NLP) in addition to emoticon sentiment classification. We also propose an analytical framework of affective computing based on the support vector machine (SVM), a machine learning method for the sentiment analysis of the public opinion on Weibo.For the application of technical analysis of HFT, we proposes an HFT analysis model based on the results of affective computing which come from the traditional financial measurement model, combined with sentiment analysis variables and time series analysis of HFT data. In addition, the elements and the framework of this model are analyzed, and an empirical test is performed. This paper also elaborates the framework design of software system, data flow and operation mode of this model, and software implementation.The study of this paper reveals the conduction of stock market sentiment in HFT and the information processing method of opinion on Weibo and provides a new reference for correlation analysis techniques.更多还原