When considering data sets characterized by a large number of instances, the computational time required to apply Genetic Algorithms for generating Fuzzy Rule-Based Classifiers increases considerably, mainly due to the fitness evaluation. Another important problem associated to these kinds of data sets is an undesired increase of the obtained model complexity. These two problems can be addressed by using Instance Selection techniques, which aim to obtain a representative subset of input data with a lower size with respect to the original set, while maintaining or even improving the classification accuracy for new input data. The aim of this study is to analyze a wide range of Instance Selection techniques together with a Genetic Fuzzy Rule-Based Classification system, namely the Fuzzy Association Rule-Based Classification Model for High-Dimensional Problems, in order to discover which reduction method or family of methods outperforms the others. The results on 36 different Instance Selection methods show that a particular family of them is very promising, since the complexity of the obtained models is significantly decreased while accuracy is maintained, in comparison with the original model without Instance Selection. This provides a first experimental framework for further developments on this type of methods, which seems to be very appropriate for Instance Selection on classification problems addressed by applying evolutionary learning of linguistic Fuzzy Rule-Based Classifiers as a kind of pre-processing that can help to significantly decrease the complexity of the obtained models.
A case study on the application of instance selection techniques for genetic fuzzy rule-based classifiers
MARCELLONI, FRANCESCO;
2012-01-01
Abstract
When considering data sets characterized by a large number of instances, the computational time required to apply Genetic Algorithms for generating Fuzzy Rule-Based Classifiers increases considerably, mainly due to the fitness evaluation. Another important problem associated to these kinds of data sets is an undesired increase of the obtained model complexity. These two problems can be addressed by using Instance Selection techniques, which aim to obtain a representative subset of input data with a lower size with respect to the original set, while maintaining or even improving the classification accuracy for new input data. The aim of this study is to analyze a wide range of Instance Selection techniques together with a Genetic Fuzzy Rule-Based Classification system, namely the Fuzzy Association Rule-Based Classification Model for High-Dimensional Problems, in order to discover which reduction method or family of methods outperforms the others. The results on 36 different Instance Selection methods show that a particular family of them is very promising, since the complexity of the obtained models is significantly decreased while accuracy is maintained, in comparison with the original model without Instance Selection. This provides a first experimental framework for further developments on this type of methods, which seems to be very appropriate for Instance Selection on classification problems addressed by applying evolutionary learning of linguistic Fuzzy Rule-Based Classifiers as a kind of pre-processing that can help to significantly decrease the complexity of the obtained models.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.