Premium
A segregated genetic programming for bioprocess modelling with outliers
Author(s) -
Wu Yanling,
Lu Jiangang,
Sun Youxian
Publication year - 2008
Publication title -
asia‐pacific journal of chemical engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.348
H-Index - 35
eISSN - 1932-2143
pISSN - 1932-2135
DOI - 10.1002/apj.207
Subject(s) - outlier , estimator , a priori and a posteriori , computer science , genetic programming , mathematical optimization , feature (linguistics) , genetic algorithm , noise (video) , nonlinear system , simple (philosophy) , algorithm , data mining , artificial intelligence , mathematics , machine learning , statistics , philosophy , linguistics , physics , epistemology , quantum mechanics , image (mathematics)
Genetic programming (GP) is often used to model a complex nonlinear system. Nevertheless, if the training data obtained from an industrial process are corrupted by large noise or outliers, the simple GP and GP based on least squares estimator usually cannot come up with an acceptable solution. To overcome this problem, a novel robust GP based on M‐estimator is proposed. Moreover, these cut‐off parameters of the estimator play a crucial role in degrading the effects of outliers. Usually an optimal value of the cut‐off parameter exists but without a priori knowledge of the training data, it is difficult to define it. So a segregated GP using two different cut‐off parameters is proposed to solve this problem. The novel feature of this approach is that the algorithm can perform multi‐directional search on the whole problem space for different cut‐off parameters, so that it can get mixed information from different directional searches and has more chance to find an acceptable solution. In addition, the proposed approach is less sensitive to the values of the cut‐off parameters and performs almost as good as a GP with an ideal cut‐off parameter. Copyright © 2008 Curtin University of Technology and John Wiley & Sons, Ltd.