Prediction of dissolved oxygen in aquaculture based on gradient boosting decision tree and long short-term memory network: A study of Chang Zhou fishery demonstration base, China

Abstract In order to further improve the prediction accuracy of dissolved oxygen (DO) in aquaculture, a prediction model of DO is proposed by combining the gradient boosting decision tree (GBDT) with the long short-term memory network (LSTM). GBDT is used to select the characteristic factors with high influence on DO. Based on the Keras deep learning framework, the LSTM model is established, and the cross-validation grid optimization algorithm is used to optimize the LSTM parameters to predict DO. The model is applied to verify DO in the standard pond of Jintan fishery base, Jiangsu province, China. The experimental results show that mean square root error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) of the model are 0.197, 0.299, and 0.092, respectively. The evaluation indexes are better than other comparison prediction models. It shows that the model has good predictive ability and generalization ability, which can meet the actual needs of modern aquaculture.