| As an important part of the financial market,the stock market plays a important role in the entire financial field.For a long time,whether in the academic field or in the field of stock investment,there are many people who are full of great interest in the prediction of the direction of the stock market.In the past,when people predicted the trend of stocks,most of them started from the perspective of the stock market itself,using the historical price of securities prices,trading volume or company-related income indicators to analyze the future price changes of stocks,seldom from economic news,financial self-media articles,stock market reviews and other text information to analyze the future price changes of stocks.With the popularization of the Internet,especially the mobile Internet,the speed of information dissemination continues to increase.Investors can easily use computers or mobile phones to obtain financial information from the Internet to assist themselves in making investment decisions.The impact of the stock market and how to use these text information to analyze and predict the changes of stock prices have very important practical significance.The research of this paper is based on the text information of financial news,from the perspective of text mining to analyze the impact and prediction of financial news on stock prices.Different from structured data such as technical indicators and fundamental indicators that are common in the market,text information is unstructured data.First,the collected financial news needs to be processed.This article first uses the Python third-party library jieba to segment each news text collected,and convert each text information into a word list,and next use the natural language processing tool Doc2 vec to represent the text information in the form of a vector,and then use the random forest algorithm to perform feature selection on the quantified text features,so as to obtain the final features of the model.This paper defines stock forecasting as a two-class model,that is,using the current text features to predict the rise and fall of the stock market the next day,and the forecast target is the CSI 300 Index.Finally,the performance of three machine learning models,support vector machine,XGBoost,and neural network on the classification task was compared.The results show that after adding text information into the model,the model’s target prediction performance is improved,which proves that the text information effectiveness.Among all machine learning models,XGBoost performs the best. |
