Cloud Resource Load Prediction Based on EEMD-LSTM

With the rapid advancement of cloud computing, accurate load prediction has become essential for optimizing system performance. The nonlinear and non-stationary characteristics of cloud resource load data have increased the complexity of prediction tasks. To address this challenge, this paper proposes a multi-input multi-output load prediction model based on Ensemble Empirical Mode Decomposition (EEMD) and Long Short-Term Memory (LSTM) networks. The model reduces data complexity through EEMD decomposition while effectively capturing the temporal dependencies in the sequence data using LSTM. Experiments conducted on publicly available cluster datasets from Alibaba demonstrate that the EEMD-LSTM model significantly outperforms traditional models, such as LSTM, Recurrent Neural Network (RNN), and Multi-Layer Perceptron (MLP), in predicting CPU and memory usage, as evaluated by metrics including Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and the Coefficient of Determination (R2).