基于自然语言处理的舆情分析和股价涨跌预测系统

Public Opinion Analysis and Stock Index Prediction System Based on Natural Language Processing

金融科技的快速发展拓展了证券市场预测股票行情的途径,传统的方法是通过股票的历史价格和特征因子进行预测,这种基于历史量价的预测方法很难反应公众情绪对股市的影响。基于深度学习的自然语言处理技术能够发掘文本数据的深度特征,从网络上的散乱信息中找到特殊线索,为金融市场中舆情的定量分析提供了解决方案。本系统以网络上关于上市公司的原始新闻资讯作为样本,以真实的股价涨跌情况作为标签,分别对样本数据进行清洗、文本向量化等预处理,设计出代表短线、中线和长线的三种数据集,搭建基于深度学习模型BERT的金融情感预测系统(BERT-FS),经过训练和评估之后,判断股票涨跌的AUC值最高可达79.48%.

The rapid development of financial technology has expanded the way to predict the stock price in the financial market. The traditional method is to predict the stock market through the historical price and characteristic factors. This method based on the historical price is difficult to reflect the impact of public sentiment on the stock market. Natural language processing technology based on deep learning can explore the depth characteristics of text data, find special clues from the scattered information on the network, and provide a solution for the quantitative analysis of public opinion in the financial market. This system takes the original news information about the listed companies on the Internet as samples, and takes the real stock price fluctuation as labels. The sample data are preprocessed by cleaning and text vectorization, and three kinds of data sets representing short-term, medium-term and long-term are designed. The financial sentiment prediction system(BERT-FS) based on the deep learning model BERT is built, which is trained and evaluated, the AUC value of judging the rise and fall of the stock can reach up to 79.48%.