A hybrid model for the prediction of dissolved oxygen in seabass farming

Perch is a relatively valuable aquatic product with high economic value. Dissolved oxygen follows a complex, dynamic and non-linear system. To solve the problems of low prediction accuracy and poor generalization ability of traditional dissolved oxygen prediction methods, a dissolved oxygen hybrid prediction model for perch culture water quality based on principal component analysis and pathfinder optimization algorithm is proposed in this paper. Firstly, the key influencing factors affecting the dissolved oxygen of bass were extracted by PCA to eliminate redundant variables and reduce the data dimension and complexity. Then the PFA optimization algorithm is used to automatically optimize the key parameters of GRU neural network to obtain the optimal parameter combination. Finally, a combined prediction model based on PCA-PFA-GRU is constructed to predict the dissolved oxygen in perch culture water quality. The MSE, MAE, RMSE and R2 are 0.010, 0.060, 0.100 and 0.983, respectively. The simulation results show that the proposed PCA-PFA-GRU model has a small fluctuation of prediction error and high prediction accuracy. In conclusion, the proposed model has good prediction accuracy and generalization and has achieved excellent prediction effect in short-term prediction to avoid huge losses, reduce growth risks and promote the development of fishery modernization.