Stock Time Series Analysis And Forecast Based On Internet Investor Sentiment

The traditional financial theory based on the efficient market hypothesis holds that financial asset prices fully reflect all available information.Behavioral finance emphasizes the irrational component of investor behavior,and believes that in addition to the intrinsic value of stocks,investor sentiment that reflects investors' psychological expectations will have an important impact on stock prices.Therefore,starting from the investment subject and studying the fluctuation of the stock market on the basis of behavioral finance theory,it is an effective way.The development of online media provides a direct source of emotional metrics.Obtaining investors' opinions and expectations from stock reviews has become an important means of financial analysis.Unstructured stock reviews pose certain challenges to the extraction of investor sentiment.In order to solve this part of the problem,this paper uses the word embedding technique and association rule method to construct and extend the sentiment dictionary,and uses the sentiment classification algorithm to predict the stock market expectations contained in the comments.Judging,and based on this,built the investor sentiment index.Further,this paper explores the relationship between network sentiment and stock market by means of the constructed sentiment index.The results of correlation analysis and causality test indicate that the sentiment index has certain predictability for stock returns.Finally,compared with the existing stock forecasting methods,this paper attempts to improve the accuracy of stock price time series forecasting from two directions.On the one hand,relying on the investor sentiment of the extracted lyric data,this paper incorporates emotion as an external feature into the predictive model.On the other hand,this paper starts with the time series model itself,combines the advantages of neural network and econometric model to construct the NARX-GARCH model,and fully explores the information contained in the price series.The experimental results show that Internet investor sentiment has guiding significance for stock market volatility,and stock forecasting method based on investor sentiment has obvious advantages.