Virtual machine (VM) consolidation emerges as a key solution to improve energy efficiency by dynamically migrating VMs to fewer physical machines (PMs) and powering down idle servers. However, aggressive consolidation risks violating service-level agreements (SLAs). This paper proposes an energy and SLA aware VM consolidation algorithm combining static thresholds, Local Regression (LR), Median Absolute Deviation (MAD), and Interquartile Range (IQR) methods for precise host overload prediction, thereby making more reasonable decisions on virtual machine migration and integration. The proposed method has been compared with other algorithms through simulation experiments. The results show that our algorithm has better overall performance. While maintaining considerable energy consumption, the SLAV value was reduced by up to 23.0% and by an average of 16.9%, and the energy-SLA violation (ESV) was reduced by up to 21.5% and by an average of 14.4%.