Sentiment Analysis of News on the Stock Market

Abstract

AbstractMedia sentiment in the stock market affects investors’ behaviors and the operation of the stock market. Meanwhile, the development of data mining and artificial intelligence makes it possible to study massive stock market news on the Internet. In order to study media sentiment in the stock market, this paper collected the stock market news of 42 constituent stocks of China Securities 100 index in 2018 from Sina News. Firstly, we constructed financial news sentiment classifier based on word2vec and support machine vector (SVM). Then, we created three new daily sentiment indexes of individual stock by using the classification results. Finally, linear regression models were built to explore their relationship with the logarithmic trading volume, turnover, yield and volatility of individual stocks. The accuracy of financial news sentiment classifier was 86%, and the regression models could explain the influence of stock market media sentiment on some indicators of stock market to some extent. It was concluded that the financial news sentiment classifier based on word2vec and SVM works well. In addition, choosing the appropriate method to convert the word vector into the text vector has a great impact on the model effect and the simple average method performed better in this paper. At the same time, media sentiment has effects on the stock market. The impacts of negative and positive sentiment on logarithmic trading volume and turnover are positive. The positive impact of neutral sentiment on turnover is weaker than that of positive and negative sentiment. KeywordsMedia sentimentStock marketSentiment analysisWord2vecSupport vector machineLinear regression