Abstract: The methods of standard deviations, maximizing deviations, entropy weight and their improved methods were applied to solve the objective weight determination problem of multivariate indexes in performance evaluation. It made the weight determination selection more comprehensive. The conflict of the indexes was introduced to take account of the correlation between indexes. The case study was implemented in VOIP network. The samples used to compute weights were obtained by simulation. Six weight computing methods were compared for selecting more objective weight determination method. The results show that Criteria Importance Through Intercriteria Correlation （CRITIC） method can get better objective weights.

Key words: network management, performance evaluation, weight, Criteria Importance Through Intercriteria Correlation (CRITIC) method, standard deviation method

摘要： 针对目前网络性能评估中多元指标之间权重难以客观确定的问题，将标准差法、离差最大化法、熵权法以及它们的改进算法应用于客观权重确定，使客观权重确定方法的选择更加全面完备。改进算法引入指标间冲突性概念，考虑指标间相关性对客观权重确定的影响。以VOIP网络为例，采用仿真方法获取权重计算样本，为了选取更加客观的权重确定方法，将六种权重计算方法进行了比较。测试结果表明，基于指标相关性的指标权重确定方法（CRITIC）计算出的权重具有较强客观性。

关键词: 网络管理, 性能评估, 权重, 基于指标相关性的指标权重确定方法, 标准差法