ѧ
JOURNAL OF SOFTWARE
1999 10 12  Vol.10 No.12 1999



ģ缰ʱзеӦ
ܴ⡡ű޴ȫ
ժҪһ͵ģģ.ģͲҪרҵ֪ʶָͨģ.ÿӦһģÿģĺΪһԱйѧϰѵ.ģʱзеӦΪͨ봫ͳʱзԼǰ緽ĶԱȣ˵ģЧ.
ؼʡģ,ģϵͳ,,,ʱз.
ͼšTP18
Fuzzy Neural Network and Its Applications in Time Series Analysis
ZHOU Chun-guang1 ZHANG Bing1 LIANG Yan-chun2 HU Cheng-quan1 CHANG Di1
1(Department of Computer Science Jilin University Changchun 130012)
2(Department of Mathematics Jilin University Changchun 130012)
Abstract A new fuzzy neural network model is presented in this paper. In the new model, the fuzzy rules are generated through the competitive classification of samples without the aid of domain experts. One class of samples forms a fuzzy rule. The consequent part of each fuzzy rule is a feed forward neural network trained using the corresponding class of samples. The application of the fuzzy neural network to the analysis of time series is taken as an example to demonstrate the effectiveness of the fuzzy neural network through the comparison with the conventional analysis of time series and neural network methods.
Key wordsFuzzy neural network, fuzzy system, samples, competitive classification, time series analysis.
ģϵͳfuzzy systemǻģϵۣԸģģʶģģƺģ.ĿǰģϵͳԶơϢ˹ܡͼʶԼѧѧѧȷ涼õ˹㷺Ӧã1.ģϵͳصֱܹӵرʾ߼ʺֱӻ߼֪ʶнǿ߼.ǣģȷȽѣͨҪרҵ֪ʶָ.
˹磨artificial neural networkܹͨѧϰݱ֪ʶ˿Լ֪Ϣ֮⣬нǿĸ.һЩģͼ㷨ѾΪ첢ﵽӦˮ׼.ܿơģʽʶ𡢻Ӿʶ֪ʶȷ涼ȡ˽չ˹.ǣе֪ʶͨȨֵģͷ2,3. 
Ŷģϵͳо룬໥Ĺϵ𽥸ı4,5.ģϵͳ໥ںϵģ磨fuzzy neural networkĲ.ģһڹҵơʱзȷõ˹㷺Ӧãչ˹Ӧǰ.68
1 ģ
11 ͵ģģ
Ƚϵ͵ģУ
(1) Hiroyuki Okadaڽڷģģͣ7ݣͼ1ʾ.SigmoidͽڵԽڵ㹹.ͨר֪ʶģģͣõģȷнڵӼȨֵõSigmoidԪʾSͻZȺSigmoidԪʵȺķзģõȷϵͳķϢѵ.ģ͵ģרҸܹרҵ֪ʶģ⣬ṹѧϰ.ǣ绥ṹȨֵȷרҵ֪ʶר֪ʶĻȡΪѣʱרҵ֪ʶҲһ.

ͼ1ڽڷģģ
(2) Cai Yaling3ֲͬͽڵ㹹ɵģͣ6.Ϊ㡢ȡСͲ㣬ֱڵ(TRAN-fuzzy neuron)ȡСڵ(MIN-fuzzy neuron)ͽڵ(SUM-fuzzy neuron)ɣʹȺ.ͲֱӦ.ѧϰУÿѧϰһȡСڵ㣬㷨ȷȺĲԼڵȨֵγһ.ͨѧϰȷȡСĽڵ.ܹѧϰټ䣬Ҫר֪ʶָ.ǣѧϰ϶ʱϴ󣬾ʶ϶.
1.2 һ͵ģģ
1.2.1 ģ
ģģͲҪרҵ֪ʶָͨģ.ÿӦһģÿĺΪһԱѧϰѵǰ.ģ4㲻ͬܵĽڵ㹹ɣͼ2ʾ.

ͼ2ģģ
Cÿģʽ
IF X1 is L1j AND X2 is L2j AND ... AND XN is LNj THEN Y=NETj(X1,X2,...,XN)
LijΪӦXiĵjԱNETjΪjĺӳ亯i=1,...,Nj=1,...,C.
ĵ1ΪԱȺ.ÿXiΪCԱÿһڵӦһԱڵΪȺֵ.ڵĴ亯Ϊ
O1ij=Lij(xi).
2ΪȡС.ڵ1ڵֵȡС˲õĵǿ(firing strength).ڵĴ亯Ϊ



FSiΪiĵǿȣi=1,...,C.
3Ϊ.ÿĺֶӦһǰ.ΪX=(X1,X2,...,XN)Աѧϰѵ.뱾ĵǿΪò.ڵĴ亯Ϊ

NETiΪiĺӳ亯i=1,...,C.
4Ϊģ.òԸķзģģ.ڵĴ亯Ϊ

1.2.2 
ΪȷģǰԱȶѧϰо.蹲MT1,T2,...,TMΪTk=(Tk1,Tk2,...,TkN)k=1,...,M.·о.
1 ռͶCӡSi=(Si1,Si2,...,SiN)i=1,...,C.
2 ÿÿӼСDikֵSiǶTkʤߣTkǵiĳԱ.ÿֻһʤߣÿֻCеĳһ.
3 еϺ󣬶ÿӸݾʤ,ʽ.

ЦΪѧϰʣȡС,SiΪTKʤ.
4 ѵ,ת2.
ھжӽʹھĹͬƶ.ξȶھõ.Ϊÿһʤߣÿһ.,ѧϰΪC.
1.2.3 Ⱥȷ
CѧϰÿӦһģ.ģǰԱȺ(4)ͼ3ʾκ״ɲa,b,c,dȷ.ԱLijĲʽȷ.

ͼ3Ⱥ


Xȼ.

1.2.4 ģѵ
Ŵ㷨ѵ8,9.ÿѧϰڱѧϰѵһҪѵ.
1.2.5 ʶĴ
ѧϰΪC࣬NάÿһάӦCԱCNԱϣCN.NϴʱCCNϵһС.ʵᷢʶйĵǿȾΪ㣬еĹû.ھʶI=(I1,I2,...,IN)IÿӵľD*i=1/D*iΪĵǿȣģΪ

þʶӦӣģľȷļ̶.ԽԽƫ򣬹ļ̶ԽС.ۺйĽķзģɵõһܴȫֵӶʶ.
2 ʵ
ģģʱзϣһЩķȽ֤Ч.ʱзϵͳʶʹõĲģУARMA(auto regressive moving average)Զع黬ƽģͣ10Ҫһ֣ARMAģ͵ĽȽϸӣʵӦгʹAR(auto regressive)ģ.
ǰ緽ǰʱԤеģʽΪѧϰӦѵѧϰлȡ֪ʶӶԤ.ǰ緽Ҫǵ϶ʱѵ.
ĵʵʹ人19781198712µƽݣ10ݣ120.ʵõģ4򣬺12룬12㣬㵥ԪΪ30.ǰ12Ϊ룬ǰһԤ⣬õһ.120ݿԲ108ģʽǰ90ģʽΪѧϰΪ.ÿŴ㷨ѵ5 000.ģıȽϼͼ4ʾ.

ͼ4ģıȽ
ǰ緽ʹþ12ڵ㡢1ڵ㡢40ڵ磬90ѧϰѵв.ͳʱзʹAR12ģͣ12Ϊ룬ǰһԣõһ.AR12ĲȷΪ
Xt=0.728Xt-1-0.246Xt-2-0.080Xt-3-0.209Xt-4-0.203Xt-5+0.018Xt-6-0.375Xt-7+0.052Xt-8+0.007Xt-9-0.114Xt-10+0.043Xt-11-0.073Xt-12.
3ַĲԽԱͼ5ʾ.ʵԿģʱз,ȡ˽ϺõЧ.

ͼ5Խ
3 
ģģͲҪרҵ֪ʶָͨľģ.ÿģĺΪһǰ磬Ŵ㷨ԱѧϰѵҽϺõؽ˾ʶĴ.ģʱзнӦãȡ˽ϺõЧ.ͨ봫ͳʱзǰ緽ĶԱȣ˵ģʵԺЧ.
ڱģģУӸȷԼȺصҪһо.
*оõȻѧ͹ҽż֪ʶ̿ʵʽ.
߼飺ܴ,1947,ʿڣʿʦҪоΪ磬ͼŴ㷨.
ű,1972,,ҪоΪŴ㷨ͼģʽʶ.
޴,1953,ʿڣʿʦҪоΪ磬ͼŴ㷨.
ȫ,1957,ڣҪоΪ簲ȫܣϢϵͳ.
,Ů1972,̣ҪоΪС任ͼʶ.
ͨѶϵ:ܴ⣬130012,ִѧѧϵ
ߵλܴ⡡űȫϡִѧѧϵ130012
޴ִѧѧϵ130012
E-mail: cgzhou@mail.jlu.edu.cn
ο
1Luo R F, Shao H H, Zhang Z J. Fuzzy-neural-net-based inferential control for a high-purity distillation column. Control Engineering. Practice, 1995,3(1):3140
2Wang De-liang. Pattern recognition: neural network in perspective. IEEE EXPERT Intelligent System & Their Applications, 1993,8(4):5260
3Kim Hongbong, Nam Kwanghee. Object recognition of one-DOF tools by a back-propagation neural network. IEEE Transactions on Neural Network, 1995,6:484487
4Satoru Isaka. On neural approximation of fuzzy system. In: Proceedings of INCNN'92, Vol1. New York: IEEE, 1992. I263I268
5Jokinen Petri A. On the relations between radial basis function networks and fuzzy system. In: Proceedings of INCNN '92, Vol1. New York: IEEE, 1992. I220I225
6Cai Yaling, Hon Keung Kwan. A fuzzy neural network for fuzzy classification. In: Proceedings of ASME SCI'94. Wuhan: Press of Huazhong University of Science and Technology, 1994. 894899
7Okada Hiroyuki et al. Initializing multilayer neural network with fuzzy logic. In: Proceedings of INCNN'92, Vol1. New York: IEEE, 1992. I239I244
8Xu H Y, Wang G Z, Baird C B. A fuzzy neural networks technique with fast back propagation learning. In: Proceedings of INCNN'92, Vol1. 1992. I214I219
9ܴ⣬ű.Ŵ㷨ѵǰеӦ.С΢ͼϵͳ,1996,17(6):5458
(Zhou Chun-guang, Zhang Bing et al. Genetic algorithm and its application in training feedforward neural networks. Chinese Mini-Micro Systems, 1996,17(6):5458)
10ӣŵ.ʱзĹӦ.人:ѧ,1992
(Yang Shu-zi, Wu Ya et al. Engineering Applications of Time Series Analysis. Wuhan: Press of Huazhong University of Science and Technology, 1992)
1998-10-09յԭ,1999-01-18յ޸ĸ
