An immune algorithm based fuzzy predictive modeling mechanism using variable length coding and multi-objective optimization allied to engineering materials processing

Chen, Jun and Mahfouf, M. (2008) An immune algorithm based fuzzy predictive modeling mechanism using variable length coding and multi-objective optimization allied to engineering materials processing. In: Granular Computing, 2008. GrC 2008. IEEE International Conference on , 26-28 August 2008, Hangzhou, China.

Documents
An immune algorithm based fuzzy predictive modeling mechanism using variable length coding and multi-objective optimization allied to engineering materials processing
In this paper, a systematic multi-objective fuzzy modeling approach is proposed, which can be regarded as a three-stage modeling procedure. In the first stage, an evolutionary based clustering algorithm is developed to extract an initial fuzzy rule base from the data. Based on this model, a back-propagation algorithm with momentum terms is used to refine the initial fuzzy model. The refined model is then used to seed the initial population of an immune inspired multi-objective optimization algorithm in the third stage to obtain a set of fuzzy models with improved transparency. To tackle the problem of simultaneously optimizing the structure and parameters, a variable length coding scheme is adopted to improve the efficiency of the search. The proposed modeling approach is applied to a real data set from the steel industry. Results show that the proposed approach is capable of eliciting not only accurate but also transparent fuzzy models.
[img]
[Download]
[img]
Preview
PDF
An_Immune_Algorithm.pdf - Whole Document

409Kb

Official URL: http://dx.doi.org/10.1109/GRC.2008.4664729

Abstract

In this paper, a systematic multi-objective fuzzy
modeling approach is proposed, which can be regarded
as a three-stage modeling procedure. In the first stage, an
evolutionary based clustering algorithm is developed to
extract an initial fuzzy rule base from the data. Based on
this model, a back-propagation algorithm with momentum
terms is used to refine the initial fuzzy model. The refined
model is then used to seed the initial population of an
immune inspired multi-objective optimization algorithm
in the third stage to obtain a set of fuzzy models with
improved transparency. To tackle the problem of
simultaneously optimizing the structure and parameters, a
variable length coding scheme is adopted to improve the
efficiency of the search. The proposed modeling approach
is applied to a real data set from the steel industry.
Results show that the proposed approach is capable of
eliciting not only accurate but also transparent fuzzy
models.

Item Type:Conference or Workshop Item (Presentation)
Additional Information:In this paper, a systematic multi-objective fuzzy modeling approach is proposed, which can be regarded as a three-stage modeling procedure. In the first stage, an evolutionary based clustering algorithm is developed to extract an initial fuzzy rule base from the data. Based on this model, a back-propagation algorithm with momentum terms is used to refine the initial fuzzy model. The refined model is then used to seed the initial population of an immune inspired multi-objective optimization algorithm in the third stage to obtain a set of fuzzy models with improved transparency. To tackle the problem of simultaneously optimizing the structure and parameters, a variable length coding scheme is adopted to improve the efficiency of the search. The proposed modeling approach is applied to a real data set from the steel industry. Results show that the proposed approach is capable of eliciting not only accurate but also transparent fuzzy models.
Keywords:artificial immune systems, fuzzy predictive modeling, multi objective optimisation
Subjects:G Mathematical and Computer Sciences > G700 Artificial Intelligence
Divisions:College of Science > School of Engineering
ID Code:2803
Deposited By:INVALID USER
Deposited On:07 Jul 2010 10:37
Last Modified:13 Mar 2013 08:41

Repository Staff Only: item control page