Optimization of prediction model of dissolved oxygen in industrial aquaculture

Dissolved oxygen affects the growth status of fishes directly in aquaculture, so a prediction model to determine the future changing trend of dissolved oxygen was set up. When the predicted values of dissolved oxygen were below the safety value, the farmer can start oxygen increasing machine in advance to maintain the safety of fishes. The proposed dissolved oxygen prediction model was based on the least squares support vector regression (LSSVR) model with chaotic mutation to improve the estimation of distribution algorithm (CMEDA) to find optimal parameters (γ and σ) of LSSVR. Because these two parameters can significantly affect the performance of LSSVR, the other three parameter optimization methods, that means, particle swarm optimization (PSO) algorithm, genetic algorithm (GA) and the traditional LSSVR, were used to compare with CMEDA algorithm. The mean absolute percentage errors of the prediction results of four models were 0.32%, 1.27%, 1.98% and 2.56%, respectively. The CMEDA-LSSVR model has a higher prediction accuracy and more reliable performance than the other models. In order to make farmers use prediction model conveniently, a dissolved oxygen prediction system GUI based on Matlab was designed. Farmers download the history data from remote monitoring system by web browser as training data and testing data, the prediction results of different time would be calculated and displayed on the GUI. The prediction model was used in Yangzhong, Jiangsu Province, China, and it performed well. It helps farmer to make decision and reduce aquaculture risks.