Improved LDA by using Distributing Distances and Boundary Patterns
Ali Yaghoubi1, Hamid Reza Ghaffar3
1Ali Yaghoubi, M.Sc., Department of Engineering, Islamic Azad University, Ferdows Branch, Ferdows, Iran.
2Dr. Hamid Reza Ghaffari, Department of Engineering, Islamic Azad University, Ferdows Branch, Ferdows, Iran.
Manuscript received on January 02, 2014. | Revised Manuscript received on January 04, 2014. | Manuscript published on January 05, 2014. | PP: 143-147 | Volume-4 Issue-6, January 2014. | Retrieval Number: F2486014615/2015©BEIESP
Open Access | Ethics and Policies | Cite
© The Authors. Published By: Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: One of the statistical methods of class discriminant is linear discriminant analysis. This method, by using statistical parameters, obtain a space which by using available discriminating information among class means does classification act . By using distributing Distances, extended analysis linear discriminant to its heteroscedastic state. At this state ,to make classes more separating of available separating information among covariance matrix classes including classes mean is using. In this article ,because of using new scattering matrices which are defined based on boundary and non- boundary patterns, classes overlapping in Spaces which obtains has been reduced . On the other hand ,using new scattering matrices brings about increasing classification rate so, the done experiments confirm improvement of classification rate.
Keywords: Boundary linear discriminant analysis, Boundary and non-boundary patterns, CHernoff criteria, linear discriminant analysis.