Exploring the methodology of port throughput prediction based on SSA-BP neural network

This paper proposes a Sparrow Search Algorithm（SSA） optimized Back Propagation（BP） neural network model for port throughput prediction, which overcomes the drawbacks of BP neural network such as slow convergence, low accuracy, and local optimum trapping. The model considers the instability and nonlinearity of port throughput forecasting and enhances the prediction accuracy. The paper uses the port throughput and related data of Xiamen Port's direct economic hinterland from 2000 to 2021 for training and simulation, and compares the model with three other models: BP neural network, Particle Swarm Optimization（PSO）-BP neural network, and Whale Optimization Algorithm（WOA）-BP neural network. The results demonstrate that the SSA-BP neural network model has higher accuracy and stability in port throughput prediction. The maximum absolute value of relative percentage error is 1.95%, the minimum value is 0.11%, the prediction error is within 2%, and the model outperforms the other three models in terms of evaluation indicators and convergence speed.