针对活性群体领导现象的格兰杰因果推断  

Granger causality inference method based on active group leadership

作　　者：周立方 万亚平[1,2] 欧阳利军 何志爽[3] ZHOU Li-fang;WAN Ya-ping;OUYANG Li-jun;HE Zhi-shuang(School of Computer Science,University of South China,Hengyang 421001,China;Hunan Provincial Base for Scientific and Technological Innovation Cooperation,China Hunan Provincial Science and Technology Department,Hengyang 421001,China;School of Information Engineering,Hunan Mechanical and Electrical Polytechnic,Changsha 410199,China)

机构地区：[1]南华大学计算机学院,湖南衡阳421001 [2]湖南省科技厅湖南省医疗大数据国际科技合作基地,湖南衡阳421001 [3]湖南机电职业技术学院信息工程学院,湖南长沙410199

出　　处：《计算机工程与设计》2023年第6期1844-1852,共9页Computer Engineering and Design

基　　金：国防科技创新特区基金项目(17-163-15-XJ-002-002-04);湖南省2020年度创新型省份建设专项抗击新冠肺炎疫情应急专题基金项目(2020SK3010);湖南省教厅重点基金项目(17A185)。

摘　　要：现有可变时间延迟格兰杰因果推断虽能较有效地发现时间序列间以任意时间延迟影响着的因果关系,但在活性群体领导现象的因果关系发现上还略有不足。对此,提出针对活性群体领导现象的可变放宽时滞格兰杰因果推断。通过对CCF正峰负偏现象进行发现,运用放宽准则重新确定新的最佳固定时滞并代入可变时滞格兰杰因果推断方法中,推断出两时间序列的因果关系。在金体美鳊鱼群觅食的实验中运用此方法,实验结果表明,在金体美鳊鱼群中,相较可变时间延迟格兰杰因果推断方法,该推断方法的效果提升了近50%。Although the existing Granger causality inference with variable time lag can effectively find the causality affected by any time lag between time series,it is still slightly insufficient in the causality discovery of active group leadership.In this regard,a variable relaxed-lag Granger causality inference based on the phenomenon of active group leadership was proposed.The positive peak and negative deviation of CCF were found,and the new optimal fixed lag was re-determined using the relaxation criterion and substituted into the variable-lag Granger causal causality method,so as to infer the causal relationship between the two time series.This method was used in the feeding experiment of Notemigonus Crysoleucas.Experimental results show that effects of this inference method are improved by nearly 50%compared with the variable-lag Granger causality inference method in Note-migonus Crysoleucas.

关 键 词：活性群体 领导现象 格兰杰因果推断 可变时滞格兰杰因果推断 动态时间规整 最佳固定时滞 可变放宽时滞格兰杰因果推断 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]