首页|基于模糊系统的定性与定量知识的综合集成

基于模糊系统的定性与定量知识的综合集成

扫码查看
综合集成法已被广泛应用于处理复杂系统相关问题,其核心是人机结合、从定性到定量的迭代求解,然而如何描述定性与定量知识,如何有效融合定性与定量知识仍是亟须解决的问题。模糊系统模拟了人脑推理过程,既可以利用专家的定性知识,也能够从数据中学习模糊规则,使用规则映射的方式实现对不确定性问题的系统决策。将模糊系统引入定性与定量知识的描述理解与融合过程,提出了一种基于模糊系统的可解释性综合集成法。该方法从定性和定量两个角度分别获取知识,再将两种知识综合集成,形成模糊规则库,完成模糊系统建模。该方法可以有效地将专家经验和数据学习相结合,增强模型的可解释性,提高复杂系统决策过程的鲁棒性和科学性,有望成为未来综合集成法研究的一种实现途径,更好地解决现实世界复杂系统的相关问题。
Qualitative and quantitative knowledge of metasynthesis based on fuzzy system
The metasynthesis has been widely used to solve complex system problems,and its core is the combination of human and machine,from qualitative to quantitative iterative solving.But how to describe and integrate qualitative and quantitative knowledge effectively is still the urgent problem to be solved.The fuzzy system simulates the reasoning pro-cess of human brain.It can not only use the qualitative knowledge of experts,but also learn fuzzy rules from the data,and use the way of rule mapping to realize the system decision of uncertain problems.By introducing the fuzzy system into the process of description,comprehension and fusion of qualitative and quantitative knowledge,the interpretable metasyn-thesis based on fuzzy system was proposed.Knowledge was obtained from quantitative and qualitative perspectives,and then the two kinds of knowledge were integrated to form a fuzzy rule base and completed the fuzzy system modeling.This method effectively combines expert experience with data learning,enhances the interpretability of the model,and im-proves the robustness and scientificity of the decision-making process of complex systems.This method is expected to be a realization method for the research of integrated method in the future,so as to better solve the problem of system com-plexity in the real world.

fuzzy systemmetasynthesiscomplex system probleminterpretability

陈德旺、刘俐俐、赵文迪、欧纪祥、孙艳焱、郑楠

展开 >

福建理工大学交通运输学院,福建 福州 350118

福建省北斗导航与智慧交通协同创新中心,福建 福州 350118

中国科学院自动化研究所多模态人工智能系统全国重点实验室,北京 100190

模糊系统 综合集成法 复杂系统问题 可解释性

2024

智能科学与技术学报

智能科学与技术学报

CSTPCD
ISSN:
年,卷(期):2024.6(4)