Intelligent Investment Algorithms Based On Stock Influence Relationship Network

With the development of economy,the practical participation in investment has become an increasingly urgent need of the public.In recent years,the rapid development of artificial intelligence technology has attracted more and more researchers and investors to try to use artificial intelligence technology intelligently and invest automatically.Quantitative investment,a new type of investment based on computer technology,has emerged in the financial field.Quantitative investment is mainly based on the assumption that history repeats itself and price prediction is made by analyzing historical price data of financial products or relevant news data.However,the financial market itself is a huge asymmetric game system,the existing technology and investment means pay too much attention to the price volatility of financial products and ignore the impact of the game relationship between each other on the price volatility.Based on this background,this paper proposes an intelligent investment algorithm based on the stock influence relation network,aiming at using the game relationship between individual stocks to predict the future trend of stock prices.At present,there is no perfect modeling scheme for the relationship between stocks.This paper proposes a model of Stock Influence Relationship modeling based on recurrent neural Network,and constructs Stock Influence Relationship Network(SIRN)based on this model.Based on the delayed co-occurrence probability of price fluctuation trend between two stocks,this model describes the effect probability of different stocks.This paper constructs SIRN in Chinese and American stock markets respectively.The experiment found that the influence relationship in the stock market was basically subject to the distribution of ?,only a small number of stocks had a strong influence relationship,and the influence relationship between most stocks was not significant.Through the analysis of the complex network structure of SIRN,it is found that the industry often presents a one-way influence relationship.Furthermore,in order to apply the influence relationship between stocks to intelligent investment,this paper proposes GF(Game Factor),a stock Factor that describes the degree to which stocks are affected by other stocks,and a voting strategy based on the influence relationship network.In this paper,different stock price trend models are tested in Chinese stock market and American stock market respectively.The experimental results show that the voting strategy performs better for Chinese stockmarket and worse for American stock market.LSTM with GF added has the best recall rate and F1,showing the best experimental performance on the whole.This shows that GF can effectively improve the prediction accuracy of the intelligent investment algorithm.Finally,an automatic investment engine is designed and implemented to transform the stock price trend model into a real intelligent investment algorithm.The real market simulation experiment on Chinese stock data shows that GF can effectively improve the yield of LSTM and SVM.It is proved that stock influence relation is important to improve the return rate of intelligent investment algorithm.