Key Techniques And Algorithms Of Cloud Computing Resource Load Forecasting

Cloud computing resource load forecasting is an important function module in the cloud computing system.Its prediction value for the future time is the key to effectively planning and running the cloud computing system.With the mass multiplication and complexity of cloud computing data,the traditional linear prediction model can't guarantee the accuracy of the nonlinear changing cloud computing resources in the short-term load forecasting.Although the neural network is suitable for the prediction of nonlinear time series,it use s the gradient descent method to calculate the threshold and weight,which lead es to the neural network easy to fall into the local optimal solution and reduce the prediction accuracy.Although scholars has introduced optimization algorithms such as particle swarm optimization in neural networks,the performance of the model is affected by the shortcomings of these intelligent algorithms.At the same time,the traditional analysis training data set method can't effectively analyze the change characteristics of load sequence.That affectes the accuracy of prediction.This paper presentes a cloud computing resource load forecasting model composed of wavelet packet transform,chaotic sine cosine whale optimization algorithm and multilayer perceptron neural network,which makes up the shortage of the existing prediction model.First,the prediction model is used the wavelet packet to decompose the predicted cloud computing resource load sequence.From the analysis of each frequency band,we can grasp the fluctuation of the resource load in each time point,so that the neural network can predict the change of the negative load in the future time more accurate ly.Then,aiming at the slow convergence rate and prematurity of the basic whale optimization algorithm,the information communication strengthening mechanism and the chaotic sine cosine mechanism are introduced into the whale algorithm.The chaotic sine cosine whale optimization algorithm is proposed.Through information exchange and strengthening mechanism,the information exchange of artificial whale population is enhanced.That avoides the individual population falling into local optimum.The chaotic sine cosine mechanism can reduce blind spots and speed up convergence.The improved whale algorithm lays the foundation for optimizing the MLP neural network and improving the performance of cloud computing resource load forecasting model.Finally,the multi-layer perceptron neural network optimized by improved whale algorithm is used to calculate the subsequence prediction values of each training sample set,and they are processed by addition transformation to obtain the normalized prediction results.In this paper,the simulation experiment is conducted under the Matlab software platform,and the cloud computing resource load data actually measured by a certain operator in Guangdong Province is used for prediction.This article sets up multiple experimental groups,and compares and analyzes different models.Through experiments,we can draw conclusions:(1)The wavelet packet decomposition c an effectively extract the sub-sequences of different frequency bands,thereby improving the prediction accuracy;(2)The chaotic sine cosine whale optimization algorithm can improve the multilayer perceptron neural network well.(3)Comparing with other predictive models,this method achieves better predictive results.