Research On Stock Price Prediction Method Based On Wavelet Analysis And Quantification Of Investor Sentiment

As an important part of the financial market,the stock market has always been the focus of financial investors.For investors,obtaining excess returns from the stock market is always their constant pursuit,which requires investors to make correct judgments on the trend of stock prices,so the prediction of stock prices since the birth of the stock market has been a hot spot in the academic and industry circles.In recent years,with the application of deep learning methods in the financial field,deep learning-based predictive models represented by LSTM models(long short-term memory neural network models)have emerged in the industry and have shown good prediction capabilities.In this study,the LSTM model was used to predict the rate of return of several assets,and in order to improve the prediction accuracy of the model,the data input of the neural network model was optimized from the aspects of data processing and factor quantification,and a quantitative strategy was constructed to simulate the prediction effect of the real market.Firstly,in terms of data processing,the wavelet transform method is used to reduce the noise of the data to improve the information quality of the input data.In terms of factor quantification,the crawler technology combined with text analysis method is used to directly quantify investor sentiment and construct a daily investor sentiment quantitative index.Secondly,using the historical return data and quantitative index of investor sentiment after wavelet transformation as the input characteristics of the neural network,the LSTM model is constructed to predict the positive and negative returns of the next trading day,and the research results prove that the investor sentiment and wavelet noise reduction processing have a significant improvement effect on the prediction effect of the neural network model.Finally,the quantitative timing strategy constructed by this study takes the prediction model as the core to construct the research target,and the quantitative timing strategy constructed in this study has strong advantages by comparing with other timing strategies and showing it in the backtesting interval.In summary,this study finds that China’s stock market,like most stock markets,is a non-efficient market.Wavelet noise reduction can remove the noise of stock market information to a certain extent,improve data efficiency,and be more conducive to neural networks to discover the relationship between data.Investor sentiment also has an improving effect on the prediction of returns,market sentiment is one of the indispensable influencing factors for studying stocks,and the direct quantification method based on the online forum data in this thesis is better than the indirect sentiment quantification method using market proxy indicators.On this basis,this study puts forward some suggestions for various investors and online information platforms on rational investment,quantitative investment and public opinion supervision.