Fig 1.
Comparison of the initial population.
(a) Random Initialization. (b) Good point set Initialization.
Fig 2.
Dynamic change diagram of the proportion of discoverer sparrows.
Fig 3.
The structure of the MSMSSA-LSTM model.
Fig 4.
The flowchart of the MSMSSA-LSTM algorithm.
Fig 5.
The 1-minute trend of the closing prices of two futures in 2021.
Table 1.
The results of basic statistical analysis.
Table 2.
The correlation coefficient of rebar and hot coil futures.
Table 3.
ADF unit root test.
Table 4.
Results of residual series stationarity test.
Table 5.
The search range for target parameters.
Table 6.
The optimization results of the objective function.
Fig 6.
Box-plot of the optimization results of the two algorithms.
Fig 7.
Convergence curves of the best optimization results of the two algorithms.
Fig 8.
Optimization iterative process of target parameters.
Table 7.
Optimization iterative process of target parameters.
Fig 9.
Comparison of the predicted value and the real value for MLP.
Fig 10.
Comparison of the predicted value and the real value for RNN.
Fig 11.
Comparison of the predicted value and the real value for LSTM.
Fig 12.
Comparison of the predicted value and the real value for GRU.
Fig 13.
Comparison of the predicted value and the real value for SSA-LSTM.
Fig 14.
Comparison of the predicted value and the real value for MSMSSA-LSTM.
Table 8.
Comparison of six models evaluation indexes.