论文题目：基于 LDA主题模型的舆情情感分析的个股交易策略设计
论文类型：交易策略设计
学科专业：金融专硕
学位申请人：刁驰
指导教师：龚秀芳（副教授）
摘要
随着互联网的飞速发展和用户人数的急剧增长，股民在论坛上的评论情绪
倾向在很大程度上反映了股民对某股票价格的未来走势的看法与判断，也影响
着股市涨跌。与此同时，市场非有效性理论也说明了股票价格不能完全反应股
票价值，即投资者的情绪和股票价格走势存在一定的相关性。由此可知，如何
有效地挖掘出投资者对股市的态度和观点，对股市预测有很大参考意义。
本文通过探究股民评论情感极性对未来股票价格的影响，在云端建立包含
月度股评数据情绪倾向特征在内的股价预测模型，为广大中小投资者的决策提
供参考。传统学者基于股评情感分析的研究大多都是事后研究，由于市场的快
速变化和不确定性，这种时间差可能影响传统预测模型结果不佳。本文将模型
完全设在云端，依托于云原生技术和 LDA主题模型，可以实现对股评数据集的
主题提取和情绪极性分析，同时配合使用 Python爬虫技术获取的负面新闻数据
集，构建出一种基于上述指标的个股交易策略，该策略是为中小投资者和投资
机构进行服务的，是一种短期的投资策略。通过对股票价格涨跌进行预测，来
实现提高收益率的目的，在模拟盘中，选取多时间段、不同公司进行模拟测试，
“跟踪”策略能够带来平均 3.75%的超额收益率，反转策略能实现平均 2.91%
的超额收益率，并能够在利好事件到来时快速反应进入市场同时也能在消极事
件期间快速撤出，测试结果证明该模型在预测准确度和误差方面的性能相比过
往研究均有提升。
II

关键词：LDA主题模型；股票价格预测；文本分析
III

Abstract
With the rapid development of the Internet and the rapid growth of the number of
users, the sentiments of investors' comments on the forum largely reflect their views
and judgments on the future trend of a certain stock price and also affect the rise and
fall of the stock market. At the same time, the market inefficiency theory also shows
that stock prices cannot fully reflect stock values, that is, there is a certain correlation
between investor sentiment and stock price trends. Therefore, how to analyze
netizens' attitudes and opinions quickly and efficiently on the stock market has great
guiding significance for stock market forecasting.
This paper try to explore the impact of investor sentiment polarity on future stock
prices, and establishes a stock price prediction model in the cloud that includes the
emotional tendency characteristics of monthly stock review data to provide a
reference for the decision-making of small and medium investors. Most of the
research based on sentiment analysis of stock reviews by traditional scholars is post-
event research. Due to the rapid changes and uncertainty of the market, this time
difference may affect the poor results of traditional forecasting models. In this paper,
the model is completely set up in the cloud, relying on cloud native technology and
LDA topic model, it can realize topic extraction and sentiment polarity analysis of
stock review data set, and at the same time cooperate with the negative news data set
obtained by using Python crawler technology to construct a kind of The individual
stock trading strategy based on the above indicators is a short-term investment
strategy for small and medium investors and investment institutions. By predicting the
rise and fall of stock prices, the purpose of increasing the yield is achieved. In the
simulated market, multiple time periods and different companies are selected for
simulated testing. The "tracking" strategy can bring an average excess return rate of
3.75%. The strategy can achieve an average excess return of 2.91%, and can quickly
react to enter the market when a positive event arrives and quickly withdraw during a
negative event. The test results prove that the performance of the model in terms of
prediction accuracy and error is improved compared to the past.
IV
。。。以下略