Research On Stock Market With External Information And Restrictive Conditions

With the continuous development of Chinese stock market,financial data appeared the growth of explosive type and multiple noises,and showed characteristics of complexity and diversity.Using single internal factors of the stock market data to do securities analysis gradually can not meet the application needs,and cannot make full use of stock data to analyze stock market.Therefore,we use the neural network to study various forms of external information in the stock market,the main content of the work is as follows.Firstly,this paper establishes a random perturbed noise layer to deal with random fluctuations in the stock market external information interference,based on the adaptability,self-learning abilities and robustness of neural networks.Random perturbed noise layer replacing the previous way of adding noise data with simulating stock market volatility risk,writing a new hidden layer to make the training model obtain the random disturbance resistance ability.Then the training model can handle the stock market random perturbance,at the same time maintain accuracy.Secondly,an exemplary model is built by using the unstructured text data in the external information.According to the natural language processing technology under the restriction of internal factor data acquisition.,a model combining Word2vec algorithm with the training network of word vector-feature data is established,as well as a word frequency statistical model based on TF-IDF.Effective characteristics of stock market are extracted from unstructured data to assist the analysis and prediction of stock market trend,It provides an idea to solve the problem of limited access to internal data when stock market trading tends to be high frequency.In this paper,the external information and restrictions existing in the stock market are targeted studied,and the results are consistent with the expectations.This indicates that deep learning can not only be used as an algorithm to analyze and predict stock data,but also create more possibilities and ideas for further research.