Stock Price Forecasting Based On Sentiment Analysis And Combinatorial Model

As an important part of the national economy,the stock market can not only help investors to formulate reasonable stock investment strategies and effectively avoid stock investment risks,but also promote the healthy development of the financial market by revealing its fluctuation law through scientific methods.Domestic and foreign research on its quantitative forecasting methods has been relatively mature,but there are still some problems,such as the singleness of influencing factors,coarse-grained research level,etc.,how to effectively improve the accuracy and efficiency of stock forecasting results,still need further research.Therefore,this paper designs a stock price prediction method based on sentiment analysis and PCA-LSTM combination model.The specific research contents are as follows:In terms of feature selection,on the premise that a single stock trading index was originally selected as the characteristic factor affecting stock fluctuations,this paper incorporated the index of text sentiment value to increase the diversity and effectiveness of the influencing factors.In the calculation of stock text sentiment value,we draw on the relevant literature of some scholars and use TF-IDF algorithm to form the exclusive sentiment dictionary of individual stocks,and design the text quantization rules based on sentiment unit.According to the comparison of the quantitative empirical results of the comment text,it is found that the sentiment analysis based on the extended sentiment dictionary can improve the accuracy of the sentiment classification results.In order to verify the feasibility of text feature sentiment analysis for stock price prediction,the correlation coefficient and significance level of text sentiment value and stock price are calculated,which provides reliable theoretical support for the model construction in the following paper.In terms of model construction,this paper combines sentiment analysis,PCA static feature extraction and LSTM dynamic time series prediction to construct an optimization model,and applies it to stock price prediction.In the first part,the PCA method is used to extract the static features of the comprehensive variables that integrate the stock trading indexes and individual stock sentiment values.The dimensions of the index variables are reduced without changing the original information,and the principal component comprehensive variable matrix is generated.In the second part,the variables after dimensionality reduction are input into the LSTM predictive neural network,which greatly improves the operating efficiency of the model while ensuring the accuracy of the model.In the aspect of empirical comparison,in order to better evaluate the performance of stock prediction based on sentiment analysis and PCA-LSTM combination model,the stock price prediction model based on sentiment analysis and LSTM and the stock price prediction model based on trading indicators and LSTM model are constructed.By comparing the fitting effect,overall evaluation error,running time and daily absolute error of the three experiments,it is proved that the prediction model proposed in this paper based on sentiment analysis and PCA-LSTM model has higher prediction accuracy and operation efficiency.Finally,the conclusions are summarized,and further research directions and ideas are put forward to provide reference suggestions for optimizing the stock forecast model.