Radial Basis Function Neural Network Model for Dissolved Oxygen Concentration Prediction Based on an Enhanced Clustering Algorithm and Adam

In fishery aquaculture, water quality directly determines the economic benefits of aquatic products, and dissolved oxygen is an important factor affecting quality. To accurately grasp trends variation in oxygen, a concentration forecasting model based on enhanced clustering algorithm Adam with radial basis function neural network (ECA-Adam-RBFNN) proposed. An (ECA) combining K-means ant colony optimization introduced place random selection to determine center positions hidden layer units. If number points too high, will be overfit, whereas if it low, sudden changes appear results. Once centers have been determined, (RBF) width calculated from maximum distance avoid two extreme cases RBF that are peaked or flat. The recursive least squares (RLS) obtain connection weights output layer. iteratively differentiate objective adjust values, while adaptively varying learning rates for these three types parameters. Finally, improved applied prediction aquaculture. experimental results show under identical conditions, compared long short-term memory (LSTM) network, backpropagation (BPNN), traditional support vector regression (SVR) model, autoregressive integrated moving average (ARIMA), K-MLPNN (K-means muhilayer perceptron networks), SC-K-means-RBF achieves significant reductions mean absolute error (MAE), percentage (MAPE) root square (RMSE) as evaluation indicators.