基于改进差分进化的智能综合评分法  

An intelligent comprehensive scoring approach based on improved differential evolution algorithm

作　　者：王荣笙 王晓刚 龙艺璇 吕宜生[2] 袁志明 WANG Rongsheng;WANG Xiaogang;LONG Yixuan;LYU Yisheng;YUAN Zhiming(Scientific and Technological Information Research Institute,China Academy of Railway Sciences Corporation Limited,Beijing 100081,China;State Key Laboratory for Management and Control of Complex Systems,Institute of Automation,Chinese Academy of Sciences,Beijing 100190,China;Signal and Communication Research Institute,China Academy of Railway Sciences Corporation Limited,Beijing 100081,China;Traffic Management Laboratory for High-Speed Railway,National Engineering Research Center of System Technology for High-Speed Railway and Urban Rail Transit,Beijing 100081,China)

机构地区：[1]中国铁道科学研究院集团有限公司科学技术信息研究所,北京100081 [2]中国科学院自动化研究所复杂系统管理与控制国家重点实验室,北京100190 [3]中国铁道科学研究院集团有限公司通信信号研究所,北京100081 [4]高速铁路与城轨交通系统技术国家工程研究中心高速铁路行车调度实验室,北京100081

出　　处：《智能科学与技术学报》2025年第1期64-76,共13页Chinese Journal of Intelligent Science and Technology

基　　金：中国铁道科学研究院集团有限公司科研开发基金项目(No.2024YJ030);北京市自然科学基金项目(No.4254110);中国国家铁路集团有限公司科技研究开发计划项目(No.Q2024Z009)。

摘　　要：针对综合评价问题中的异常评分导致的综合评分结果有失公允问题,提出基于改进差分进化的智能综合评分法,通过剔除异常评分提升综合评分结果的公平性。首先,从异常评分产生的根本原因出发,定义综合评价问题的公平性指标及目标函数,分析问题约束条件和异常评分判定条件。其次,提出基于改进差分进化的智能综合评分法,即在传统差分进化算法的框架上做出以下3点改进:提出基于评分值权数的实数编解码方式;采用比重法将评分值权数等式约束写入目标函数,将综合评价问题转化为无约束优化问题,提升算法求解效率;以公平性指标为基础,设计评分值权数优化问题知识,采用启发式种群初始化方法提升算法前期收敛速度和评分值权数结果的求解质量。最后,算例结果表明,所提出的智能综合评分法能自动判定权数较小的评分值为异常评分并予以剔除,最终在13 s内给出公平性更高的综合评分结果,有效提升了综合评分结果的公平公正性与科学合理性。In response to the issue of biased comprehensive evaluation results caused by anomalous ratings in comprehensive evaluation problems,an intelligent comprehensive scoring approach based on improved differential evolution algorithm had been proposed,which enhanced the fairness of comprehensive evaluation results by eliminating anomalous ratings.Firstly,considering the causes of abnormal scores,the fairness index and objective function were defined for the comprehensive evaluation problem.The constraint and the criterion of abnormal scores were also analyzed.Secondly,an intelligent comprehensive scoring approach had been proposed based on improved differential evolution algorithm,which introduced the following three improvements to the traditional differential evolution algorithm framework:a real encoding and decoding method was developed based on the score weights.The weighting method was used to write the constraints of the score weights into the objective function.As a result,the comprehensive evaluation problem was transformed into an unconstrained optimization problem to improve efficiency.The problem-specific knowledge of score weights was designed based on the fairness index.A heuristic population initialization method was employed to speed up the algorithm convergence in the early stage and the solution quality of the score weights.Finally,the case study results had demonstrated that the proposed intelligent comprehensive scoring approach could automatically identify ratings with smaller weights as anomalous and eliminate them,ultimately providing a more fair comprehensive evaluation result within 13 seconds,effectively improving the fairness,impartiality,and scientific rationality of the comprehensive evaluation results.

关 键 词：综合评价 综合评分法 差分进化 进化计算 专家评审 异常评分 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程] N945.16[自动化与计算机技术—控制科学与工程]