Abstract
Pre-processing of classification data can be helpful regardless of the type of classifier. The objective of this pre-processing step is to achieve a high degree of separation among classes before the classifier is trained or tested. This results into a trace ratio problem which is difficult to solve. Methods such as Linear Discriminant Analysis (LDA) have already been used for the solution of this problem by turning it into a simpler yet inexact problem. Also, in classical LDA, the covariances of different classes are assumed to be similar, which is not the case in real-world problems. In this paper, a class-dependent approach to finding the linear transformation is proposed. This method solves the trace ratio problem directly and also removes the requirement of similar covariance matrices. While giving good results, the method is computationally expensive. To reduce the computational cost while maintaining the benefits of the class-dependent method, a multi-objective formulation is proposed and solved using NSGA-II. Simulation results show great improvement in classification using various classifiers.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsPreview
Unable to display preview. Download preview PDF.
References
Moeinzadeh, H., Mohammadi, M.-M., Akbari, A., Nasersharif, B.: Evolutionary-Class Independent LDA as a Pre-Process for Improving Classification. In: Genetic and Evolutionary Computation Conference (July 2009)
Duda, R.O., Hart, P.E., Stork, D.: Pattern Classification. Wiley, Chichester (2000)
Belhumeur, P., Hespanha, J., Kriegman, D.: Eigenfaces vs. Fisherfaces: recognition using class specific linear projection. IEEE Trans. Pattern Anal. Machine Intell. 19(7), 711–720 (1997)
Coello, C.A.C., Lamont, G.B., van Veldhuizen, D.A.: Evolutionary Algorithms for Solving Multi-Objective Problems, 2nd edn. Springer, Heidelberg (2007)
Deb, K., Pratap, A., Agarwal, S., Meyarivan, T.: A Fast and Elitist Multiobjective Genetic Algorithm: NSGA-II. IEEE Trans. Evolutionary Computation 6(2), 182–197 (2002)
Fisher, R.: The use of multiple measurements in taxonomic problems. Annals of Eugenics 7, 179–188 (1936)
Eiben, A.E., Smith, J.E.: Introduction to Evolutionary Computing, 2nd printing. Springer, Heidelberg (2007)
Witten, H., Frank, E.: Data Mining: Practical Machine Learning Tools and Techniques, 2nd edn. Elsevier Science & Technology, Amsterdam (2005)
Lozano, M., Herrera, F., Krasnogor, N., Molina, D.: Real-Coded Memetic Algorithms with Crossover Hill-Climbing. Evolutionary Computation Journal 12(3), 273–302 (2004)
Wang, H., Yan, S., Xu, D., Tang, X., Huang, T.: Trace Ratio vs. Ratio Trace for Dimensionality Reduction. In: IEEE Conference on Computer Vision and Pattern Recognition, pp. 1–8 (2007)
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2009 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Khalidji, M., Moeinzadeh, H., Akbari, A., Raahemi, B. (2009). LDA Pre-processing for Classification: Class-Dependent Single Objective GA and Multi-objective GA Approaches. In: Corchado, E., Yin, H. (eds) Intelligent Data Engineering and Automated Learning - IDEAL 2009. IDEAL 2009. Lecture Notes in Computer Science, vol 5788. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-04394-9_56
Download citation
DOI: https://doi.org/10.1007/978-3-642-04394-9_56
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-642-04393-2
Online ISBN: 978-3-642-04394-9
eBook Packages: Computer ScienceComputer Science (R0)