To improve the accuracy of software defined networking （SDN） traffic prediction, an SDN traffic prediction model based on an adaptive spatiotemporal network is proposed. This model captures the spatial correlation of SDN traffic by using an adaptive graph convolutional neural network, captures temporal variation trends through gated recurrent units, and introduces an autoregressive module in response to the highly dynamic nature of SDN traffic. The experimental results show that the proposed SDN traffic prediction method can identify more traffic features compared to existing baseline models and demonstrates superior prediction performance.