The Stock Trading Strategy Design Based On GRU-improved LSTM Gate Control Long And Short-term Memory Network

In recent decades,financial quantification has emerged and has matured rapidly.For financial institutions such as funds,investment institutions are becoming increasingly unsatisfied with passively building the average return of the portfolio income market,and the importance of active quantitative strategy portfolios is increasing,which requires the introduction of active stock investment fund management models.At present,there are many active quantitative investment strategies in China's stock fund investment market.The algorithms used are very different,such as SVM,random forest,RNN circular memory network,and so on.This article looks at this trend,using the already emerging LSTM-GRU long-term and short-term memory network model in the field of financial stock investment to build a set of active investment stock strategies based on natural language processing and other fields.Comparison of widely used models such as SVM and RNN.In theory,compared to SVM relying solely on kernel functions for higher-order mapping and classification of data,neural network algorithms such as RNN and LSTM-GRU are better in principle and more suitable for processing financial stock data.Later,through multiple comparisons of single-phase AUC,multi-phase AUC,and modeling time,it was finally found that the LSTM-GRU long-term and short-term memory network has a good accuracy rate.The LSTM-GRU algorithm was used to build a trading strategy based on the Shanghai and Shenzhen 300 index constituent stocks,and parameter adjustment In combination with the neural network,it finally achieved a significant performance over the benchmark index CSI 300 in the long term.The conclusion of this article is that the research results provide a certain quantitative strategy reference for finainstitutions to build active stock portfolios.