Self-similar traffic prediction model based on decomposition-optimized LSTM for space-ground integrated network

In the space-ground integrated network, various services such as user communication data, sensor collection information, and multimedia transmission will generate a large number of self-similar data streams. The self-similarity of the data flow in the space-ground integrated network can cause network instability problems such as network congestion, increased latency, queue overflow, and increased packet loss rate.To address these challenges, this paper proposes a self-similar traffic prediction model based on a decomposition-optimized Long Short-Term Memory (LSTM) network to forecast service traffic in the upcoming time segments. This enables proactive network transmission planning. Firstly, a traffic decomposition model utilizing the Crested Porcupine Optimizer and Variational Mode Decomposition (CPO-VMD) is developed to decompose the original service traffic of the space-ground integrated network into multiple modal components. Next, a self-similar traffic prediction model based on the Sparrow Search Algorithm and Long Short-Term Memory (SSA-LSTM) is constructed to accurately predict each modal component, which are subsequently recombined. Simulation results demonstrate that the proposed SSA-LSTM self-similar traffic prediction model, enhanced by CPO-VMD decomposition, achieves high prediction accuracy with low computational complexity. The RMSE comparison algorithm of this algorithm is 29.21% lower than the average of LSTM prediction, EMD decomposition/LSTM prediction, VMD decomposition/LSTM prediction, and MAE is 52.66% lower on average, so the proposed algorithm can effectively predict the self-similar business flow in the space-ground integrated network.