Research And Application Of Cloud Platform Load Prediction Method Based On Width Learnin

With the development of cloud computing technology,cloud platform has been widely used in various fields.As the number of cloud platform users grows,cloud platform operators are particularly concerned about how to improve resource utilization.Prediction of cloud platform load is an effective method for dynamic resource allocation.The broad learning system is a shallow neural network with the advantage of fewer network parameters,fast training and accuracy.Therefore,this paper applies the method based on the broad learning system to predict cloud platform load.The main works are as follows.First,a broad learning system based on the sparrow search algorithm is proposed to predict the cloud platform load.This method uses the sparrow search algorithm to optimize the shrinkage factor and regularization coefficient in the broad learning system,and the optimal combination of hyperparameters obtained from the optimization is used to train the broad learning system.In the experiments,determine the CPU and memory utilization to represent the load situation of the cloud platform by the coefficient of variation method.The proposed model and other models are used to forecast the load of the cloud platform,and the results show that the proposed model can achieve better prediction accuracy than the others.Second,a functional broad learning system load prediction method is suggested to reduce the sampling interval requirement and improve the generality of the load prediction model.This method takes both smooth curves and discrete observations as inputs.The long-term timing and correlation features of the curve samples are extracted using functional coefficients.The timing information at the end of the observation sequence is extracted through the weight vector.In addition,to make the functional broad learning system work best,the sparrow search algorithm is used to optimize its hyperparameters,and a cloud platform load prediction method based on the sparrow search algorithm for the functional broad learning system is finally formulated.Ablation experiments and cloud platform load multi-step prediction experiments demonstrate the effectiveness of model improvement.Finally,a cloud platform load warning system was developed and implemented.Based on the requirement analysis and design scheme,the system was developed using Python and ECharts that achieve the integration of forecasting algorithms and the visual display of forecasting sequences.