Abstract：In order to improve the prediction accuracy of the dissolved oxygen in Litopenaeus vannamei aquaculture ponds, a dissolved oxygen prediction model based on deep belief network （DBN） model and least squares support vector regression （LSSVR） was proposed. First, the DBNs were employed using ex- traction feature vectors of time series water quality data. Then, the feature vector as a training and test set for DBN-LSSVR model training and optimization. The combinations of the best parameters were obtained automatically after the optimization, which construct the nonlinear prediction model between the dissolved oxygen and the impact factors. Finally, validation and comparative analysis of the model were carried out using the measured data in Panyu district, Guangzhou. The proposed prediction model of DBN-LSSVR had high prediction accuracy and generalization ability, and it was a suitable and effective method for predicting dissolved oxygen in intensive density Litopenaeus vannamei culture.
2017, Vol. 30 
