Research On Stock Price Forecasting Method Based On ARIMA-BP Neural Network And Sentiment Analysis

With the continuous improvement of China’s national income level,people’s concept of financial management and investment is also increasingly strengthened.As the most typical representative of financial investment,stock has the characteristics of high risk and high income at the same time,so it has been highly concerned by people.However,stock fluctuations are often accompanied by the returns of investors,so how to avoid risks in advance and improve returns has become an issue of industry research and discussion.In recent years,the research in the field of stock prediction is divided into two directions.One direction is to improve the stock prediction algorithm,or to combine different prediction algorithms,so as to enhance the accuracy of the prediction.The other direction is to analyze the news emotion tendency,to simulate people’s subjective thoughts,and then obtain the correlation between news emotion and the rise and fall of stocks,so as to achieve the purpose of prediction.However,for the volatile stock market,these two research directions are not comprehensive,and the accuracy of stock price prediction needs to be improved.Therefore,this paper combines the two methods,not only using different prediction algorithms to form a combined prediction model method,to explore the rules from the data;It also uses the method of news sentiment analysis to obtain the sentiment tendency from the text.Finally,combining the combination model and the sentiment analysis,it puts forward a new direction and research ideas,and can further prove the correlation between the sentiment and the value in the stock market fluctuations.This paper focuses on the Shanghai and shenzhen A shares in domestic stock data as an example,at the same time with sina finance and economics network news as text data,first design adding emotional dictionary,and then to improve the term weighting,and the word-sentence-paragraph-the entire article-on the same day with all of the sequence of emotions tend to value the quantitative processing,finally get the daily news emotions tend to value stocks.At the same time,the ARIMA time series model and BP neural network model were constructed respectively,and the news sentiment orientation value was fused with the key indexes of stocks to form a new data set,and the parallel combination was carried out by determining the weight.Finally,the ARIMA-BP neural network combined with financial news sentiment tendency value is constructed to predict the closing price of stocks.The experimental results show that the effect of this method on stock price prediction is better than that of the ARIMA-BP neural network combined model without sentiment analysis.This paper analyzes and forecasts the stock price and development trend from the two dimensions of numbers and words,which provides investors with a more reliable objective basis for judgment,so that investors can better avoid the stock market risks and improve the returns brought by stocks.