The Research On The REITs Index Prediction Based On Att-LSTM Model

In June 2021,infrastructure public offering REITs will be listed in China,marking the official launch of the REITs pilot program.REITs have become an important investment and financing tool.The financial and real estate properties of REITs make the REITs index vulnerable to nonlinear factors.From the perspective of economic physics,this paper first uses LLE(Largest Lyapunov exponent,LLE)and DFA(Detrended Fluctuation Analysis,DFA)to find that the REITs index has chaotic and long-range correlation;then uses the sequence forward selection algorithm based on random forest to reduce the dimension and noise of the features,and based on the LSTM model,Combined with the attention mechanism,the Att-LSTM(Attention based-LSTM)model is constructed.Compared with the LSTM model,the GRNN model and the SVM model,it is found that the Att-LSTM model is better than the LSTM model and the GRNN model whether it is REITs index or trend prediction,and SVM model;and found that the fractal characteristics of REITs index have a significant impact on REITs index prediction.Finally,the investment strategy is constructed based on the model used in this paper,and it is found that the 2-8 round strategy is optimal,and the 2-8 round strategy based on the Att-LSTM model is better than the 2-8 round strategy based on the SVM model,which can effectively improve the annual to maximize returns and reduce maximum drawdowns.