Զѧ
ACTA AUTOMATICA SINICA
2000Vol.26No.3P.378-381



ģģͼڸϵͳʶеӦ
Žաë١졡κɻ
ժҪڶģ߼ۣһڸϵͳģģģ.ģͺģ֯ṹϣ˸ģϵģȺĻ㷨.ģȣANFISģģͼСȸߣڸάݽģиΪ.ַ.
ؼʡģģֶͣӦ˲ANFIS.
FUZZY-TREE MODEL AND ITS APPLICATIONS TO 
COMPLEX SYSTEM MODELING
ZHANG JiangangMAO JianqinXIA Tian
(The Seventh Division,Beijing University of Aeronautics & Astronautics,Beijing100083)
WEI Kehui
(Academy of Engineering Physics,Chengdu610003)
AbstractBased on the binary tree structure and fuzzy logic theory,a fuzzy-tree model applied to complex system modeling is proposed in this paper.Linear models and fuzzy sets are arranged in a tree structure.A hybrid-training algorithm is used to update linear model coefficients and membership functions of the fuzzy sets.Compared with some other modeling methods,such as ANFIS,the proposed model is of less computation,higher accuracy,especially for high dimension data modeling.Simulation results demonstrate the advantages of this approach.
Key wordsFuzzy-tree model,piecewise linear adaptive filter,ANFIS.
1
ıʶԴϵͳĽģ.֮£TakagiSugenoġģģڿơԤȷȡɹӦ.Ȼÿֱ𻮷ģӿռķһЩ׿Խģģ12ݵĽģЧ.
ṹֶӦ˲3ݲöṹռ仮ģӿռ䣬仮ֵӿռά޹.ģ͵ķֶκཻϵ.
һṹģģͣΪģģ.ȿ˷ģģռ仮ֵĸԣʹ÷ֶκཻƽɣܸõرƽϵͳ.
2ģģ͵Ľṹ
¸ϵͳĽģ
y=f(x)+v,(1)
x=1x1xnTRn+1ΪyRΪvRΪ.
Ϊ޷ǿռFl(.)r(.)FFȣ0ʣ
1)tFl(t)=r(t)=0l(t)>t, r(t)>t;
2)tFFСشΨһsFʹt=l(s)t=r(s).
tFl(t)r(t)ֱʾtɵҽڵ.FеСԪسΪڵ㣬Ϊr(F).s,tFt=l(s)t=r(s)stĸڵ㣬tsӽڵ㣬s=p(t).ڵtǸڵ㣬Ҷڵ㣬l(t)=r(t)=0.ʾFҶڵ㼯.
һģģһFӦ.tFȨct=ct(0),,ct(n)TģNtȺΪt(x):Rn+10,1.ӦģΪ
(2)
Ntһģӿռt.r(F)Ϊռ䣬ڵϵģNr(F)Ϊ
(3)


ͼ1
ڽڵ㶨帨ȺΪ
(4)
ЩڵϵģNtʽƵõ
(5)
ɴ˿ɵ
(6)
գģģͽռ仮ΪҶڵӦģӿռ䣬ÿģӿռڶһ
(7)
ģ1еķģģģ͵ΪҶڵļȨƽ
(8)
ͼ1ʾռĻֹ̣б߱ʾģӿռ֮ģ.
3ģģ͵Ľģ㷨
3.1ѵ㷨
ѵݼΪ{(xi,yi),xiRn+1,yiR,i=1,2,,L}.iѵݵܵΪ
(9)(10)
ģģҪȷĲΪcttF,t,ttF,tr(F).Ĳûѵ㷨ct,tFС㷨Ԧt,tF,tr(F)÷Ż㷨.
(11)
3.2ģ
ÿڵ㶨ڵetͷֲڵect
(12)
(13)
3.3ģģͽģ
1)ʼģڵ㡪ڵȨcr(F)r(F)(x)=1.ѡttF,tr(F)ֵl(t)=-r(t),r(t)0.
2)(12)(13)ʽÿҶڵetect.
3)ҶڵпԷֲĽڵֲ棬(13)ʽɵҶڵĦt,tֲֵҶڵȨֵct,t.
4)ѵt,t.
5)(5)ʽҶڵȺҶڵȨֵct,t.(10)ʽ.et>ectת2򣬽㷨.
4
.ģģԤ-˹ʱСϵͳģĹʱ׼.µ-˹ӳ΢ַ̣4
(14)
1гģģ͵ıȽϣΪ=17y=x(t+6)n=4NDEIΪĿеı׼.
1Ƚ

ѵNDEI
ģģ5000.012 4
ANFIS25000.007
ھĹƷ55000.014
ARģͣ25000.19
ģͣ65000.06
BPģͣ65000.02
׶ʽģͣ65000.04
Ԥⷽ62 0000.55

2ģģANFISıʶȽ

ռάNDEI
ģģ100.00516
ANFIS40.00716

ɱ1ؿģģͺANFISԤ⾫ȸ.ڱ2бȽģģANFISάn=10ANFISģѧռĿnָʵn>4ʱ.ģ㷨άУάʹ.
ANFISCԣn=4Apollo 700ϵйվ1.5СʱģMatlab Mļڱ586/166ϼn=105.
5ᡡ
һָϵͳıʶģģ͡ģģͣӦ˻Ͻģ㷨.ɷԿģڸ߶ȷϵͳʱеĽģеõ˺ܺõĽ.ģȽϣַСȸߣرڸάϵͳ.
Ȼѧ.
Žա1993ҵڱպѧԶϵ1999ѧʿѧλ.оΪƹ̣оΪСϵͳʶеӦáϵͳܻʶܻƵ.
ë١1964ҵڱѧѧѧϵ1982ѧ˶ʿѧλ1985ѧʿѧλαպѧڡʿʦйԶѧ¡IEEEϵͳѧᱱֻϯҪоΪƹ.
Žգպѧоҡ100083
ë٣պѧоҡ100083
죨պѧоҡ100083
κɻݣоԺɶ610003
ο
1ݡTakagi T,Sugeno M.Fuzzy identification of systems and applications to modeling and control. IEEE Trans. Syst. Man, Cyber., 1985,15:116132
2ݡJyh-Shing Roger Jang. ANFIS: adaptive-network-based fuzzy inference system. IEEE Trans. Syst. Man, Cyber., 1993, 23(3): 665685
3ݡGelfand Saul B,Ravishankar C S. A tree-structured piecewise linear adaptive filter. IEEE Trans. Info. Theory, 1993,39(6): 19071922
4ݡMackey M, Glass L. Oscillation and chaos in physiological control systems.Science, 1977,197:287289
5ݡChiu S. Fuzzy model identification based on cluster estimation.Journal of Intelligent & Fuzzy Systems, 1994,2(3): 267278
6ݡCrowder RS. Predicting the Mackey-Glass time series with cascade-correlation learning. In:Proc.Connectionist Models Summer School, Carnegie Mellon University, 1990,117123
ոڡ1998-03-09
޸ڡ1998-07-29
