效能评估可信度的客观度量方法

摘要

针对目前装备系统效能评估可信度无法客观度量的问题,提出了一种客观度量评估可信度方法.该方法从评估数据、评估模型及单样本分类置信度3个方面描述评估可信度.在计算评估数据可信度方面,通过分析评估数据的不确定性确定评估数据的可信度;在计算评估模型可信度方面,将模式分类准确率转化为评估模型可信度度量指标;在计算单样本分类置信度方面,根据不同类别样本在高维空间的分布关系,给出了适用多数分类算法的分类置信度计算方法;进一步通过可信度传播模型将以上3个方面进行综合,得到综合评估可信度.对雷达抗干扰效能评估可信度的仿真结果表明:该方法不使用专家经验知识,能够客观度量效能评估的可信度,其结果不受评估主体影响,解决了不同专家对同一装备给出的效能评估可信度不同的问题;与现有主观方法相比,准确率提高了10％.%An objective measurement method for the credibility of evaluation from three aspects:evaluation data,evaluation model and single sample confidence is proposed to solve the problem that effectiveness evaluation credibility cannot be objectively measured for equipment systems.The credibility of evaluation data is determined by analyzing the uncertainty of the evaluation data;and the model classification accuracy is transformed into a credibility index of the evaluation model.A method to calculate the classification confidence is proposed based on the distribution of the different types of samples in a space with high dimension and the method is applicable to most classification algorithms.These aspects are then integrated through a credibility propagation model to get a comprehensive credibility.Finally,the concrete application of the method is discussed in detail by taking the credibility calculation of effectiveness evaluation of a radar anti-jamming efficiency as an example.It is found from the analysis of the application that the method can objectively measure the credibility of effectiveness evaluation without the needs of the knowledge of experts,the results are not affected by evaluation subjects,and the problem that different evaluation experts give different credibility for the same equipment effectiveness evaluation is solved.The calculation accuracy of the proposed method is 10％ higher than that of an existing subjective method.