Evaluation on Software Performance Optimization Based on Multi-objective Optimization Algorithms

Abstract

The tasks undertaken by software in various tasks include precise positioning of computational targets, utilization of hardware resources, and so on. Therefore, updating software performance is an essential and important step. However, currently, software iteration in computers is not smooth sailing. Therefore, this article aimed to study the problem of software performance optimization through the direction of objective optimization algorithms. This article used the method of evaluating parameter optimization results to conclude that multi-objective optimization algorithms have stronger computational power in models in other fields. The optimization results of the target software are statistically analyzed through web log mining, and the Chebyshev method is also used to decompose one of the classic multi-objective algorithms, MOEA/D. Finally, the optimization approach was divided into two types: multi-objective optimization algorithm and single objective optimization algorithm. Two other computers with identical configurations were selected to use these two algorithms for experiments. The multi-objective optimization algorithm was set as the experimental group, while the single objective optimization algorithm was set as the control group. After performing several software optimization operations, the results showed that the software performance optimization operation based on the multi-objective optimization algorithm improved the software running speed much higher than the control group executing the single objective optimization algorithm. The final conclusion is that in the research of software performance optimization, MOEA/D in multi-objective optimization algorithms not only defeats single objective optimization algorithms, but also is a highly efficient means in itself.