Multivariate Time Series Forecasting for Cloud Machine Utilization

The cloud machine utilization prediction using a multivariate time series model is explored in this study. The cloud machine usage data set samples are selected with varied patterns. The long sequence forecasting model Informer is investigated for the forecasting. The model results are benchmarked against LSTM, GRU, Prophet, and XGBoost across the diverse datasets. The performance is evaluated using RMSE. Experimental results show that the Prophet model performs best overall due to its ability to capture both seasonal and irregular patterns accurately. The RMSE value for Informer is 0.13, Prophet is 0.097, XGBoost is 0.096, LSTM is 0.1725, and GRU is 0.1547.