Investor Sentiment Analysis And Application Based On Deep Learning

The financial market is full of opportunities and challenges.With the rapid development of science and technology,investors can easily participate in financial markets,and more and more retail investors participate in the stock investment.Investors aim for profit and they expect to get more information in favor of stock investment decisions.While traditional stock forecasting is based on historical stock price data.This thesis describes the investor sentiment of the stock market through the sentiment analysis of the stock review text,and forecasts the stock market with the combination of historical stock price data and technical indicators.The main content and innovation of this thesis are given below.1)In view of the scarcity of Chinese stock review corpus in the financial field,this research selects the appropriate data sources by comparing the major financial information platforms.Then use 'Octopus' crawler tool to obtain Gub review text data.Finally,after data cleaning,Chinese stock review corpus is constructed by manual sorting and tagging to meet the needs of investor sentiment analysis.2)In view of the current situation of stock review emotion recognition,this thesis proposes a method of stock review emotion recognition combining attention mechanism and slice RNN.Through large-scale corpus training,this method overcomes the disadvantages of emotion dictionary classification and the inefficiency of traditional RNN recognition methods.At the same time,attention mechanism makes up for the loss of long-term dependence of the low-level network of RNN model.3)This thesis puts investor sentiment into the method of market prediction,Through web crawlers and sentiment analysis models,market investor sentiment is quantified and sentiment indicators are constructed.Combine historical market price data and construct technical indicators to verify that investor sentiment has a certain impact on stock market forecasts.Experiments show that the proposed model has certain reference significance in the prediction of short-term GEM index price trend.