A Study On Price Prediction Of Financial Products Incorporating Multi-source Data Sentiment Analysis

The financial market is the place where funds are financed,not only manipulating the inflow and outflow of funds,but also controlling the lifeline of the national economy.Only when the financial market is in good condition,the economy will be in a balanced state and provide a relatively stable environment for economic development.Financial products are the basis for the development and existence of financial markets,as well as the connecting vehicle for financial market participants.Stocks are an important part of financial products,and a correct grasp of their sentiment tendencies is of great value for accurate stock market prediction.At present,traditional sentiment analysis techniques are mainly based on sentiment dictionaries and machine learning,and sentiment analysis factors are relatively single.This study integrates multi-source data from news information,stockholders’ comments and market trading prices,combines behavioral finance and efficient market hypothesis,and explores the influence of sentiment factors on stock market volatility using a two-stage deep learning approach.The specific research contents are as follows.First of all,this research proposes a sentiment analysis model based on Convolutional Neural Networks(CNN).In order to avoid the inability of single text data to effectively capture stock market fluctuation trends and stockholders’ sentiment tendency,the thesis takes stockholders’ comments and financial news information together as text data sources.After data cleaning,integrity and standardization tests,the text information is word vectorized by Word2 vec.Taking the recent stockholders’ high concern of Xiangxue Pharmaceutical,Guizhou Maotai and Baiyunshan stocks as examples,the sentiment values of stockholders’ comments and financial news information are extracted by inputting CNN sentiment analysis model.Then,the article proposes a stock price prediction model based on sentiment analysis of multi-source data.Structured data of market trading prices,financial statements and sentiment values obtained by CNN sentiment analysis model are combined as the input of Bi-directional Long Short-Term Memory(Bi LSTM);Attention mechanism is introduced to make the model extract global temporal sequence of stock commentary text while capturing key pattern information.The introduction of Attention allows the model to capture key pattern information while extracting global timing information of stock commentary text.The experimental results show that the average absolute percentage error of the financial product price prediction model incorporating deep learning from multiple sources of data is reduced by 5.99 compared to using only a single stock index,and the results indicate that the sentiment factor of stockholders’ comments and financial news information can have a certain degree of influence on stock price prediction.