In cloud computing, efficient virtual machine (VM) management is paramount for ensuring optimal resource utilization, reducing energy consumption, and maintaining service-level agreements (SLAs). Traditional VM allocation and switching techniques often struggle with workload unpredictability, leading to performance degradation or resource underutilization. This research introduces a novel hybrid methodology combining Explainable AI (SHAP) and the proposed Secretary Bird Artificial Hummingbird Algorithm (SBAHA), a fusion of the Artificial Hummingbird Algorithm (AHA) and the Secretary Bird Optimization Algorithm (SBOA). SHAP is used for workload prediction, offering interpretable insights into feature impact, while SBAHA is leveraged for optimizing VM migration and server switching decisions based on multiple fitness parameters including power, load, scalability, resource constraints, reliability, and network performance. A round-robin-based dynamic allocation further enables scalable resource management. Experimental results demonstrate that the proposed SBAHA outperforms existing techniques such as PSG, HHM, FL, and SW, achieving superior performance with lower MAE (1.46), RMSE (8.31), and MAPE (0.02), while significantly reducing execution time and number of VM migrations. These improvements highlight SBAHA potential as a robust solution for intelligent cloud infrastructure management.