Abstract: In software-defined networking（SDN） environment, network traffic is based on the complexity of topology and temporal dynamic characteristics, which leads to traffic prediction facing the dual challenges posed by spatial and temporal characteristics.To solve this problem, a traffic prediction model based on whale algorithm（WOA） is proposed.The model improves the prediction accuracy by fusing the extraction ability of complex neural network（CNN） for spatial features and the capture ability of long short-term memory network（LSTM） for time series features, and optimizes the model hyperparameters by WOA.Finally, it is compared with classical methods such as CNN-LSTM and PSO-LSSVM.The results show that the WOA-CNN-LSTM model has a relative reduction of 80.91%,69.21% and 72.91% in MAE,RMSE and MAPE metrics compared to the CNN-LSTM model, and 40.29%,19.10% and 34.76% compared to the PSO-LSSVM,respectively.The experiments verify the good performance of the model in SDN traffic prediction and provide new ideas for traffic prediction in complex network environments.