基于熵权层次分析法的云平台负载预测

为提高对云平台负载情况的预测精度,降低预测负载值的波动性,研究非线性时间序列的预测算法。对影响云平台运行的因素进行认知和分解,提出基于熵权层次分析法的主客观评价模型。在此基础上,建立基于差分整合移动平均自回归模型(autoregressive integrated moving average model,ARIMA)和BP(back propagation)神经网络的云平台负载预测模型。实验结果表明,基于熵权层次分析法的评价模型吸收主客观赋权法的优势,能够合理准确地反映云平台的实时负载情况;基于ARIMA-BP的预测模型得出的预测结果具有更小误差的精度和波动较小的数值稳定性。

Cloud platform load forecasting based on entropy weight analytic hierarchy process

To improve the prediction accuracy of the cloud platform load and reduce the volatility of the predicted load,the prediction algorithm of nonlinear time series was studied.The factors that affected the operation of cloud platform were recognized and decomposed,and the subjective and objective evaluation model based on entropy weight analytic hierarchy process was proposed.On this basis,a cloud platform load forecasting model based on ARIMA(autoregressive integrated moving average model)and BP(back propagation)neural network was established.Experimental results show that the evaluation model based on entropy weight analytic hierarchy process absorbs the advantages of subjective and objective weight method,and can reasonably and accurately reflect the real-time load of cloud platform.The prediction results based on ARIMA-BP prediction model are more accurate and numerically stable.