Stock Oriented Network Public Opinion Information Processing And Fluctuation Trend Prediction

There is a close connection between stock related information and stock price fluctuation. If we want to use the stock related information to forecast the stock fluctuation trend, we need to consider how to quantify investor sentiment and extract events from stock related public opinion information,how to grasp the potential complex relationships among sentiment, events and stocks. Therefore, this is a focus for our research?Therefore, with the above mentioned problems, we studied stock related public opinion information processing and fluctuation trend prediction. Mainly for the following three aspects:1.The capture of stock oriented network public opinion information.There are three kinds of stock related opinion information, i.e. stock information data (e.g. stock news, stock announcement), stock market data (e.g. opening price, closing price) and stock topic data (e.g. post,blog). Stock information data can be obtained from Wind information service provider, and stock market data can be obtained from TuShare(the interface package of financial data), but for the acquisition of stock topic data, we need to design a specialized network crawler to crawl data by ourselves. Therefor we design and implement a distributed network crawler to obtain data from eastmoney guba.2. Near duplicate information removal. We observed that the post data set obtained from eastmoney guba and the news data set obtained from Wind contain a lot of repeated or near duplicated information. In the study of this paper, the types of repeated or near duplicate information are text, so we will call both of the two types of information near duplicate document for convenience. The existence of near duplicate documents can interfere with our research, so we propose a near duplicate document detection algorithm (SigNCD) based on compression technology.3. The prediction of stock fluctuation trend. The stock market movement is essentially driven by new information. Financial news and social emotion can both impact the stock market fluctuation. Here, we call the information contained in financial news as events. In this paper,we study how to extract company events and investor sentiment from the history stock oriented public opinion data, and establish a model(tensor+Ml+M2) to predict stock fluctuation trend based on tensor technology.The research contents and achievements of this paper provide some good ideas and solutions for the stock oriented network public opinion processing and fluctuation trend prediction. So our study has important research value and great application significance.