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

Abstract

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

A Study of Optimization in Deep Neural Networks for Regression

“…Supplementing average value [16,17], linear interpolation [17,18], KNN [19,20] Dimensionality reduction PCA [8], pooling layer [21], t-SNE [22], SPCA [23] Removing outliers Pauta criterion [18], EWMA [24] Feature selection PSO [1], LASSO [8,25], ASO [26], GA [27], MI [28], GRA [29], PCC [30], CCA [31] Decomposition EMD [32], EEMD [33], CEEMDAN [19,27,[34][35][36][37][38], ICEEMDAN [39], SSA [40,41], VMD [42], SVMD [43] Normalization [6,17,20,27,30,31,34,42,[44][45][46][47][48][49]…”

“…The gap is mainly based on empirical values. Different from Tsokov et al [17], Shao et al [16,19] utilized the KNN method to obtain missing values by calculating the average value of adjacent data, which is not necessary to consider the gap of missing values.…”

“…The experimental results showed that the proposed models could perform better than without removing outliers. Decomposition methods are used to process datasets, such as network traffic datasets [19], solar datasets [34], and wind datasets [39], to avoid noise in model distortion. Li et al [32] compared the performance of DNNR models using empirical mode decomposition and those without it.…”