Cloud resource prediction model based on LSTM and RBF
Na Zhang, Yifan Su, Biao Wu, Xiaomei Tu, Yuting Jin, Xiaoan Bao · 2021
In order to respond to resource usage "in advance" and to dynamically, timely and accurately schedule and allocate resources based on predicted values, a cloud resource prediction model based on Long Short Term Memory (LSTM) neural network and Radial Basis Function (RBF) neural network is proposed after investigating the current state of research on the cloud container resource prediction problem, and the parameters are trained using back propagation (BP) neural network to find the optimal combination of parameters. The LSTM can avoid the long-term dependency problem, and remembering information from very early moments is the default behavior of the LSTM without the use of a back propagation (BP) neural network. The LSTM can avoid the long-term dependency problem, and remembering information from very early moments is the default behavior of the LSTM without paying a great price for it specifically. The transformation of RBF from input space to hidden space is nonlinear, while the transformation from hidden space to output space is linear, greatly speeding up learning and avoiding the local minima problem .The experimental results show that the average error ratios of the LSTM single prediction model and RBF single prediction model are 16.38% and 19.36%, respectively, while the average error ratio of the proposed model is 4.51%, which shows that the model proposed in this paper has better prediction performance and prediction accuracy compared with other models.