Analysis and Prediction of Energy Efficiency in Cloud Computing Based on Simulated Annealing Algorithm Optimized Support Vector Machine Models

Abstract

In this study, a support vector machine hybrid model optimized based on simulated annealing algorithm is proposed for the energy efficiency prediction problem in cloud computing scenarios. By combining the global optimality-seeking property of the simulated annealing algorithm with the nonlinear modeling advantage of support vector machines, a prediction framework with adaptive parameter tuning capability is constructed, aiming to provide data-driven decision support for improving the energy utilization efficiency of cloud computing systems. The experimental results show that the model exhibits significant learning ability in the training stage, with a coefficient of determination (R²) of 0.793, indicating that the model can effectively capture 79.3% of the variant features in the dataset, with a mean absolute error (MAE) of only 0.0708, a prediction bias strictly controlled within 7.1%, and a mean value of deviation (MBE) close to the neutral value (-0.0019), which is both not The mean value deviation (MBE) is close to the neutral value (-0.0019), which does not show systematic prediction bias and maintains the dynamic balance between prediction accuracy and generalization ability. In the testing and validation session, the model still maintains robust performance when facing unknown data, with the R² coefficient maintained at 0.147, the MAE index stabilized at 0.156, which is only about 1.2 times larger than the error in the training set, and the MBE value slightly increased to 0.0096, reflecting that the model only has a controllable optimistic bias of 0.96% in the unfamiliar data environment, and this kind of error propagation characteristic across data sets validates the robustness of the algorithmic architecture. Notably, the consistency of the error magnitude between the training set and the test set (in the range of 10^-1 to 10^-2), as well as the limited fluctuation of the MBE metrics within the positive and negative intervals, together confirm that the optimization model possesses the dual mechanisms of suppressing overfitting and maintaining the stability of prediction.