Application of LSTM-GRU in green financial market trend prediction

Abstract

In view of the multi-factor coupling and non-linear fluctuation characteristics of green financial market, and the difficulty of traditional forecasting models to capture both long-term policy trends and short-term market fluctuations, this paper proposes a multi-parameter trend forecasting method for green financial market based on LSTM-GRU hybrid model. By integrating the long-term memory ability of LSTM and the lightweight computing characteristics of GRU, a joint forecasting model of green financial composite index, constituent stock return, market turnover and net capital inflow is established. Experiments show that the average absolute error of the green financial composite index prediction in real market transaction data is reduced to 0.85, the average absolute percentage error of return prediction is compressed to 5.2%, the accuracy of volume prediction is improved by 32% compared with a single model, and it is robust to green financial market mutations. By analyzing the dynamic coupling mechanism of multi-parametric indicators, the model provides high-precision and high-efficiency prediction support for investment decision-making and regulatory supervision of green financial market.