Cloud Computing Resource Prediction Based on Adaptive Selection of Multiple Machine Learning Methods

In modern cloud computing environments, accurate prediction of resource demand is critical to optimize resource allocation and reduce costs. Existing methods are often inadequate in dealing with variable workloads. In this research, we introduce an innovative combined multiple machine learning prediction approach to enhance prediction accuracy and stability. First, a comprehensive dataset is built by segmenting historical resource utilization and extracting feature vectors using dynamic variable windows. The variable window automatically adjusts its size according to the volatility, thus saving resources. Second, a number of different algorithms are used for forecasting, and the optimal method is selected to capture resource utilisation patterns and improve forecasting accuracy. Finally, the system calls the optimal algorithm from the adaptive selector for prediction based on the feature vectors. The experimental results show that the method performs well under various workloads and scenarios and significantly outperforms the traditional method, providing a theoretical basis and practical guidance for improving the efficiency and performance of cloud computing resource management.