摘要
在计算机系统性能评价模型中,指标权重分配方案是影响评价结果的重要因素。考虑到现有主流权重赋权方法存在的不足,提出兼顾主客观法的双层混合赋权模型,以尽可能地提高评价结果的可信度。该方法根据领域专家对准则层的主观评价,首先采用层次分析法给出各指标权重;然后,根据基准测试程序测试结果,对指标层采用离差最大化法确定对应准则层的指标权重,并通过混合赋权确定指标的最终权重;最后,结合标准化之后的评判矩阵确定待测系统之间的性能优劣顺序。实验表明,本评价模型能够较精确的完成评价计算机系统性能的任务。
In Computer System Performance Evaluation Model,the distribution scheme of metrics'weight is an important factor on impacting the final evaluation result.Considering the weakness of the current prevalent weighting methods,this paper proposes a double layer hybrid weighting model considering both subjective and objective methods to improve the credibility of the evaluation result as much as possible.It first computes the weights of criteria layer metrics by analytical hierarchy process according to field expert's subjective evaluation,then computes the weights of indicator layer metrics under the scope of criteria layer by maximizing deviation based on the results of benchmark,and next computes the final weights of metrics by hybrid weighting method.Finally determines the performance priorities of the candidate computer systems with the normalized decision matrices.Experiments reveal that this evaluation model is able to fulfill the task of computer system performance evaluation accurately.
出处
《微处理机》
2010年第6期114-118,共5页
Microprocessors
关键词
计算机性能评价
层次分析
离差最大化
基准测试程序
Computer Performance Evaluation
Analytical Hierarchy Process
Maximum Deviation
Benchmark