Effect of penalty function parameter in objective function of system identification

The evaluation of an objective function for a particular model allows one to determine the optimality of a model structure with the aim of selecting an adequate model in system identification. Recently, an objective function was introduced that, besides evaluating predictive accuracy, includes a logarithmic penalty function to achieve a suitable balance between the former model’s characteristics and model parsimony. However, the parameter value in the penalty function was made arbitrarily. This paper presents a study on the effect of the penalty function parameter in model structure selection in system identification on a number of simulated models. The search was done using genetic algorithms. A representation of the sensitivity of the penalty function parameter value in model structure selection is given, along with a proposed mathematical function that defines it. A recommendation is made regarding how a suitable penalty function parameter value can be determined.

[1]  K. Dejong,et al.  An analysis of the behavior of a class of genetic adaptive systems , 1975 .

[2]  David E. Goldberg,et al.  Genetic Algorithms in Search Optimization and Machine Learning , 1988 .

[3]  Mohd Azlan Hussain,et al.  Model structure selection for a discrete-time non-linear system using a genetic algorithm , 2004 .

[4]  John J. Grefenstette,et al.  Optimization of Control Parameters for Genetic Algorithms , 1986, IEEE Transactions on Systems, Man, and Cybernetics.

[5]  Goldberg,et al.  Genetic algorithms , 1993, Robust Control Systems with Genetic Algorithms.

[6]  Sheng Chen,et al.  Model selection approaches for non-linear system identification: a review , 2008, Int. J. Syst. Sci..

[7]  Zbigniew Michalewicz,et al.  Genetic Algorithms + Data Structures = Evolution Programs , 1996, Springer Berlin Heidelberg.

[8]  Mitsuo Gen,et al.  Nonlinear mixed integer programming problems using genetic algorithm and penalty function , 1996, 1996 IEEE International Conference on Systems, Man and Cybernetics. Information Intelligence and Systems (Cat. No.96CH35929).

[9]  George Kapetanios,et al.  Variable selection in regression models using nonstandard optimisation of information criteria , 2007, Comput. Stat. Data Anal..

[10]  David B. Fogel,et al.  Evolution-ary Computation 1: Basic Algorithms and Operators , 2000 .

[11]  Hishamuddin Jamaluddin,et al.  Optimum grouping in a modified genetic algorithm for discrete-time, non-linear system identification , 2007 .

[12]  Lennart Ljung,et al.  System identification (2nd ed.): theory for the user , 1999 .

[13]  Aris Spanos,et al.  Akaike-type criteria and the reliability of inference: Model selection versus statistical model specification , 2010 .

[14]  Dan Boneh,et al.  On genetic algorithms , 1995, COLT '95.

[15]  Sándor M. Veres,et al.  Structure selection of stochastic dynamic systems : the information criterion approach , 1991 .

[16]  Elisa Guerrero Vázquez,et al.  A Penalization Criterion Based on Noise Behaviour for Model Selection , 2001, IWANN.

[17]  Lennart Ljung,et al.  System Identification: Theory for the User , 1987 .

[18]  Xin Yao,et al.  Evolutionary Optimization , 2002 .

[19]  John H. Holland,et al.  Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control, and Artificial Intelligence , 1992 .

[20]  A. E. Eiben.  Genetic algorithms + data structures = evolution programs: Z. Michalewicz. Springer, Berlin, 1996, 3rd revised and extended edition (1st edition appeared in 1992), 387 pp. (hardcover), 68 figures, 36 tables, price DM 58 , 1997 .