Forecasting Model of Fishery Import and Export Trade Data Using Deep Learning Method

Fisheries import and export trade is an important component of agricultural trade and has a significant impact on national economic growth. Through the forecast of the development of fishery import and export trade, it can provide reference for the fishery management department to formulate reasonable policies and regulations to promote the sustainable development of fishery. This paper uses the import and export trade amount data in the China Fishery Statistical Yearbook as the experimental sample. In order to solve the characteristics of overfitting and insufficient generalization ability that are prone to occur in few-shot prediction, the idea of weighted moving average method is introduced for data enhancement. Through experimental simulation, optimizing the hyperparameters of the Bidirectional Long Short-Term Memory Recurrent Neural Network (BiLSTM) to build a predictive model. Input the expanded samples into the trained BiLSTM prediction model to predict the import and export trade data of fishery aquatic products. Experimental results show that this model has higher prediction accuracy and stability than commonly used models such as Gated Recurrent Units (GRU), Long Short-Term Memory (LSTM), and Back-propagation (BP) neural network.