基于组合赋权与模型自优化的能力评价  

Ability evaluation based on combination weighting and model self-optimization

作　　者：张睿 白晓露 赵娜 李吉 潘理虎[1] 陈立潮[1] ZHANG Rui;BAI Xiao-lu;ZHAO Na;LI Ji;PAN Li-hu;CHEN Li-chao(School of Computer Science and Technology,Taiyuan University of Science and Technology,Taiyuan 030024,China)

机构地区：[1]太原科技大学计算机科学与技术学院,山西太原030024

出　　处：《计算机工程与设计》2022年第8期2342-2351,共10页Computer Engineering and Design

基　　金：山西省研究生教育改革研究课题基金项目(2019JG171);教育部产学合作协同育人基金项目(201902016003、201801121011);全国高等学校计算机教育研究会课题基金项目(CERACU2019R02);太原科技大学教学改革与研究基金项目(201937);山西省高等学校教学改革创新基金项目(J2021429);山西省研究生教育改革研究课题基金项目(2021YJJG244)。

摘　　要：综合考虑评价流程中的综合性、自适应性、可靠性等因素,提出基于组合赋权与模型自优化的能力评价方法,以多级评价指标组合赋权结合改进模糊算法构建评价数据集;利用格拉姆角和场空间域转换优化序列样本信息丰富度,将粒子群优化算法、迁移学习与神经网络模型有效结合,通过模型中10个参数自适应全局寻优,构建出适用于小样本、具备自学习能力的P-TMVGG评价模型。通过实例验证了所提方法的有效性,为相关领域构建综合评价、预测、诊断体系及方法设计提供新的思路。To address the problems of comprehensiveness,adaptability and reliability in the evaluation process,a capability evalua-tion method based on combination weighting and model self-optimization was proposed.The evaluation data set was constructed by combining multi-level evaluation index weighting and improved fuzzy algorithm.Gramian angular summation field domain transformation were used to optimize the information richness of sequence samples.Meanwhile,particle swarm optimization algorithm,transfer learning and neural network model were effectively combined.Through the self-adaptive global optimization of 10 parameters in the model,a P-TMVGG evaluation model with self-learning ability was constructed,which was suitable for small samples.The effectiveness of the proposed method is verified by an example,which provides a new idea for the construction of comprehensive evaluation,prediction and diagnosis system and method design in related fields.

关 键 词：组合赋权 格拉姆角和场 粒子群优化算法 自适应 迁移学习 综合评价 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]