用于时间序列数据建模的多模态模糊认知图  

Modeling Time Series Using Multi-Modality Fuzzy Cognitive Maps

作　　者：冯国亮 卢伟[1,2] 杨建华 Feng Guoliang;Lu Wei;Yang Jianhua(School of Control Science and Engineering,Dalian University of Technology,Dalian 116023,China;Liaoning Industrial Equipment Distributed Control Technology Innovation Center,Dalian 116023,China)

机构地区：[1]大连理工大学控制科学与工程学院,辽宁大连116023 [2]辽宁省工业装备分布式控制专业技术创新中心,辽宁大连116023

出　　处：《系统仿真学报》2022年第3期543-554,共12页Journal of System Simulation

基　　金：国家自然科学基金(62073056,61876029);国家重点研发计划(2019YFB1705103);中央高校基本科研业务费专项资金(DUT20LAB129)

摘　　要：针对单一模型难以准确反映时间序列多种变化模态的问题,提出了一种基于模糊认知图的时间序列数据多模态建模方法。该方法使用随机自助法选取多个子序列,以包含各种变化模态。在各个子序列上分别建立子模糊认知图模型。使用粒计算方法对子模型进行有效融合;并分析了不同权重策略融合的性能。所建立的模型不仅可以对时间序列数据进行数值及区间预测,还具有更好的语义解释性。在公开数据集上的实验结果证明了该方法的有效性和实用性。A multi-modality modeling method for time series data based on fuzzy cognitive maps is proposed to address the problem that a single model is difficult to accurately reflect the multi-modal characteristics of time series.The bootstrap method is used to select multiple sub-sequences from the original time serieswhich contain the diverse modality in the original time series.The fuzzy cognitive map sub-models are constructed on each sub-sequencesrespectively.The formed sub-models are further merged by means of granular computing method and the merging performance with different weighting strategies is analyzed.The developed multi-modal model not only has prediction abilities at the numeric and interval level,but also has better interpretability.Experimental results on public datasets exhibit the usefulness and satisfactory efficiency of the proposed approach.

关 键 词：模糊认知图 粒计算 时间序列 模型权重 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置] TP391.9[自动化与计算机技术—控制科学与工程]