Research On Predicting Of Stock Price Patten Indomestic Stock Market Based On Complex Network And DNN

The classification and prediction of stock price fluctuation law is a very important problem in stock market research.How to describe and predict the fluctuation of stock price model has become one of th e hot issues in the field of financial research.The complex network i n stock market and the prediction of stock price fluctuation patterna re important problems in stock price research.Previous studies have u sed historical information on individual stocks to predict future tre nds in stock prices,with little regard for the movements of stocks w ithin the same market.In this paper,the information of related stock s is extracted for prediction,and the combination of complex network and deep learning is used to predict the stock price pattern.Usingthe stock closing price data from the beginning of 2010 to the end of 2019 of China's two major stock market indexes,this paper constructsa c omplex network,calculates the characteristic variables of networktop ology,and predicts the next day model of stock prices,which provide s some reference ideas for preventing financial risks and strengtheni ng macro-prudential supervision.The purpose of this paper is to use the daily closing price data of the two major stock indexes of China's stock market to obtain the price model of the stock indexes according to the time series of mult iple stocks.For each combination pattern of each stock,a complex net work is constructed,and four network topology characteristic variabl es are calculated.Variables are substituted into the model to predict the next day pattern of the stock index.The research contents and methods of this paper include(1)Const ructing a pattern network to describe the daily combination symbol pa ttern corresponding to each stock index,taking the daily combination pattern of two major stock indexes as the node of the network,and de termining the edge and weight of the network in chronological order;(2)According to the constructed directed network,the characteristic variables of network topology,including network average degree centr ality,average network strength,average shortest path length and net work proximity centrality,are calculated.(3)Four network topology c haracteristic variables are used as input variables of deep learning and machine learning models to predict stock price patterns;(4)Compa ring the prediction accuracy of deep learning and machine learning,a relatively better prediction model was obtained.The relevant conclusions obtained in this paper include(1)The p rediction accuracy of DNN algorithm in deep learning algorithm is hig her than that of machine learning algorithm selected in this paper;(2)Among the machine learning algorithms,the prediction accuracy of SVM algorithm is significantly higher than that of other algorithms.In ot her words,the generalization ability of DNN model is greater than th at of SVM classification model,and is greater than that of other wid ely used machine learning models.