Sequential Series-Based Prediction Model in Adaptive Cloud Resource Allocation for Data Processing and Security

A developed adaptive forecasting model for cloud resource allocation is presented. It employs principal component analysis on a sequence of virtual machine (VM) requests. Requests are processed to detect anomalies, and adaptive predictions are computed using EEMD-ARIMA or EEMD-RT-ARIMA methods. The choice between EEMD-ARIMA and EEMD-RT-ARIMA methods is determined by comparing the execution time values ${\mathrm {R}}_{\mathrm {{i}}}$ (sequential series test) with the threshold value ${\mathrm {R}}_{\mathrm {{t d}}}$. If ${\mathrm {R}}_{\mathrm {{i}}} \gt {\mathrm {Rtd}}$, EEMD-ARIMA is used; if ${\mathrm {R}}_{\mathrm {{i}}} \leq {\mathrm {R}}_{\mathrm {{t d}}}$, EEMD-RT-ARIMA is applied. This adaptive approach enables the selection of a prediction method based on data characteristics and resource demands. To optimize the selection of the ${\mathrm {R}}_{\mathrm {{t d}}}$ threshold, the impact on accuracy and time costs is examined. A quartile method is utilized to detect dynamic spikes, and cubic spline interpolation is employed to smooth data. EEMDRT-ARIMA-based forecasting enhances accuracy through preprocessing of dynamic spikes and adaptive method selection. Calculations of time costs indicate that this method reduces forecasting time by 1.5 times by extracting core component sequences.