Research On Key Technologies In Stock Market Forecast Based On Social Media Text

With the development of the financial industry, the stock market has influenced our social life from various aspects. And the stock market has become one of academic fields that can’t be ignored, and the research on the stock market forecast has received extensive attention. But the influence of investor sentiment on the return rate is not considered in a typical model. In order to optimize the investment strategy, this thesis proposes a fusion model based on the analysis of the social media in the form of text.Then, an improved investment decision model is designed by using fuzzy controller. The improved model can obviously improve the accuracy of short-term prediction and the yield of investment decision. The work mainly includes the following two aspects:Firstly, based on the social media text analysis, this thesis makes a short-term prediction of stock market trend. At present, there are two short-term forecasting models of stock market including using historical data and using investor sentiment. The traditional stock market trend short-term forecasting usually uses predefined sentiment analysis model to extract the prediction metrics. It predicts the short-term trend of stock price, by using the prediction index and financial time series. However, as the scarcity of the resources in Chinese sentiment analysis domain, and the difference in the stock market and the traditional sentiment analysis domain, it is difficult to forecast the trend by standard sentiment analysis module. Based on this, this thesis proposes an automatic annotation method to construct the data set, and extracts the sentiment tendency of the text through the text classification model. Finally, by using the Tianya, Sina and Baidu stock market sub section as data set, and testing 164 days SSE (Shanghai Stock Exchange) Composite Index, the improved model has a better result than the basic models.Second, on the basis of fuzzy controller, this thesis uses a variety of stock market data to establish an improved investment decision model. In the process of investment decision, it is an important process to get a new Index to ensure that the model is effective for a long time. As investor sentiment is very important in stock market forecast, so we get a new index from social media to improve the basic model. For that,we use the genetic algorithm to optimize fuzzy control rules, and encode as IS(k) by using the result of investment decision in the short term. And then, improve a basic investment decision model. According to the experiments result, this model is better than the random method. And it is better than the original investment decision model too. It proves that the combination of more indexes can improve the rate of return.