Stock Market Sentiment Monitoring And Prediction Based On Mass Text Mining

Web2.0 and mobile internet have spawned the generation of massive user content on the Internet.Users in the securities market similarly like to express their opinions and emotions on the market through social media.Therefore,social media has attracted a lot of attention from researchers and industry in recent years,but extracting from these social media is helpful for research and products.There are various difficulties in the information.In order to effectively extract the emotional information of the securities market users from social media and explore its impact on the securities market,this paper proposes a market sentiment classification algorithm for the securities market based on the characteristics of the user content data in the securities market;The problems of financial time series forecasting are presented.Risk-Return-based financial time series forecasting algorithm is proposed.Then,the forecasting model is applied to the actual trading situation,and the performance in the market is studied by means of quantitative transaction analysis.The main contents of this paper include the following three aspects:Firstly,aiming at the chaotic characteristics of user data in the Web2.0 era,this paper presents the Encoder-Decoder sentence characterization method,which maps all texts into highdimensional semantic vectors and labels datasets.Train neural network classification model,and carry out sentiment classification for mass data.The accuracy rate of classification is 81.23%,and it is compared with the accuracy of existing methods,which proves the effectiveness of the method and considers the labeling cost and the different labeling amount.Compared with the existing methods,this paper proves that the classification method of this paper has the advantage of excellent performance in a small number of annotation sets,so that emotional monitoring can be performed on the securities market through emotion classification;secondly,the analysis is faced with the analysis of traditional financial time series forecasting.The problem of forecasting only the rate of return without considering the risk and the problem of non-linear relationship between data features cannot be expressed.In this paper,the Risk-Return model is proposed by using the volatility as a measure of risk.At the same time,the short-term memory neural network is used to characterize the sequence data.Non-linear relationship,and experimentally validated Risk-Return financial time The effectiveness of the intersequence model;third,combining market sentiment data and financial market forecasting models,the prediction results form a quantitative strategy,and study the performance in the market,the results show that the model predicts high-yield-low-risk stocks to achieve accuracy 59.411%,quantitative strategies in the market achieved 13.357% annualized rate of return,and the risk indicators are lower.The contributions of this paper lie in the following three aspects.Firstly,the useroriented emotion classification algorithm of Encoder-Decoder+NN,which is based on the characteristics of user review data in the securities market,proves the effectiveness of the method and enables market sentiment monitoring.Second,the introduction of long-term and short-term memory neural networks to predict financial time series predictions.In terms of forecasting results,both profitability and risk values are taken into account.Third,the exploration of deep learning and big data in the field of financial risk has been explored and proved.The impact of sentiment on the securities market and ways to measure the impact.