LSTM Parameter Optimization Based on Improved Sparrow Search Algorithm and NSGAII

The Long Short-Term Memory (LSTM) network is a commonly used model for time series data. However, its model performance and generalization ability heavily depend on the parameter configuration. Therefore, it is necessary to study parameter optimization for the LSTM model. Sparrow Search Algorithm (SSA) holds advantages over the traditional optimization algorithms in several aspects. However, it still faces some challenges. Therefore, based on SSA, this paper introduces three strategies to improve it: Levy flight strategy, sine search strategy, and step size factor adjustment strategy. The resulting algorithm is called the Improved Sparrow Search Algorithm (ISSA). Then, we establish the ISSA-LSTM model. Additionally, to address the limitations of SSA in dealing with multi-objective problems, we introduce the Fast Non-dominated Sorting Genetic Algorithm (NSGAII) and establish the ISSA-NSGAII model. Finally, the experimental validation demonstrates the superiority of the proposed algorithm in optimizing parameters for the LSTM model.