Variable cluster analysis method for building neural network model

Abstract

To address the problems that input variables should be reduced as much as possible and explain output variables fully in building neural network model of complicated system, a variable selection method based on cluster analysis was investigated. Similarity coefficient which describes the mutual relation of variables was defined. The methods of the highest contribution rate, part replacing whole and variable replacement are put forwarded and deduced by information theory. The software of the neural network based on cluster analysis, which can provide many kinds of methods for defining variable similarity coefficient, clustering system variable and evaluating variable cluster, was developed and applied to build neural network forecast model of cement clinker quality. The results show that all the network scale, training time and prediction accuracy are perfect. The practical application demonstrates that the method of selecting variables for neural network is feasible and effective.

Keywords

variable cluster / neural network / information theory / cluster tree

References

[1] SkrzpekTNeural Network Simulation Environments [M], 1993, Massachusetts, Kluwer Academic Publishers

[2] PattersonD WArtificial Neural Networks: Theory and Applications [M], 1996, New Jersey, Prentice Hall

[3] ArbibM MThe Handbook of Brain Theory and Neural Networks [M], 1995, Boston, The MIT Press

[4] KastraM. Forecasting combined with neural networks[J]. Journal of Forecast, 1996, 15(1): 49-61

[5] WidrowB. Neural network application in industry, business and science[J]. Communcation of the ACM, 1994, 37: 93-105

[6] YanPing-fan, ZhangChang-shuiArtificial Neural Network and Computing with the Simulation of Evolution [M], 2000, Beijing, Tsinghua University Publishers(in Chinese)

[7] HornikK. Multilayer feedforward network are universal approximators [J]. Neural Network, 1989, 15(2): 359-366

[8] JedwardJ. Principal components and factor analysis: principal components(Part 1) [J]. Journal of Quality Technology, 1980, 12(4): 512-517

[9] MichaelJ, PiovosoK, KosanovichA. Applications of multivariate statistical methods to process monitoring and controller design [J]. Control, 1994, 59(3): 743-765

[10] RenRuo-en, WangHui-wenMultiple Statistic Data Analysis-theories, Methods and Examples [M], 1997, Beijing, Defence Industry Publishers(in Chinese)

[11] PalN R, BezdekJ C. On cluster validity for the fuzzy mean model[J]. IEEE Trans on Fuzzy System, 1995, 3(3): 370-379

[12] FangKai-tai, PanEn-peiCluster Analysis [M], 1982, Beijing, Geologic Publishers(in Chinese)

[13] DonaldG FObject-oriented Requirements Analysis and Logical Design [M], 1993, New York, John Wiley & Sons Inc

[14] GrahamIObject-oriented methods [M], 1993, London, Addison Welseley Publishers Ltd

[15] ShenWei, HuangWen-xi, MinPan-rongTechnology of Cement [M], 1995, Wuhan, Wuhan Industry University(in Chinese)