A Hybrid Approach by CEEMDAN-Improved PSO-LSTM Model for Network Traffic Prediction

As an important part of data management, network traffic evaluation and prediction can not only find anomalies but also judge the future trends network. To predict more accurately, a novel hybrid model, integrating Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) long short-term memory neural (LSTM) optimized by improved particle swarm optimization (IPSO) algorithm, is established for prediction. Firstly, LSTM model real-time mutation dependence constructed, IPSO applied to optimize hyperparameters. Then, CEEMDAN introduced decompose sequences raw into several different modal components containing information reduce complexity sequence. Finally, experiments shows feasibility effectiveness proposed method comparing it other deep architectures regression models. The results show that CEEMDAN-IPSO-LSTM produced significantly superior performance reduction error.