A Hybrid Machine Learning Approach to Cloud Workload Prediction Using Decision Tree for Classification and Random Forest for Regression

The dynamic nature of cloud workloads necessitates accurate predictions to optimize resource utilization, enhance performance, and ensure quality of service (QoS). Consequently, numerous researchers have developed workload prediction models to improve cloud design and deployment. These models enable timely and reliable workload forecasting, facilitating critical decisions such as resource allocation and network bandwidth management. This study proposes a hybrid learning model, termed DTCRFR, which integrates Decision Tree Classification and Random Forest Regression techniques to predict reliable workloads. The DTCRFR model operates by initially assigning a workload state to each input data point based on historical workload data and system metrics. Subsequently, the regression model refines this prediction, producing a highly accurate workload value for the classified state. The combined approach enhances prediction accuracy while reducing computational complexity, making it highly suitable for real-time applications. Empirical results validate the effectiveness of this hybrid model, demonstrating improved prediction accuracy and reduced mean-squared error (MSE) and mean absolute error (MAE). This highlights the benefit of combining classification and regression techniques to leverage their complementary strengths for more reliable and granular workload predictions. The proposed method significantly enhances resource management and system performance in diverse computational environments. By merging classification and regression, DTCRFR adds precision and subtlety to workload forecasting, providing a notable advancement in the field. This hybrid model improves both efficiency and reliability in workload prediction, marking a relevant contribution to cloud resource optimization.