PCA for improving the performance of XCSR in classification of high-dimensional problems

Abstract

XCSR is an accuracy-based learning classifier system (LCS) which can handle classification problems with real-value features. However, as the number of features increases, a high classification accuracy comes at the cost of more resources: larger population sizes and longer computational running times. In this research, we present a PCA-enhanced LCS, which uses principal component analysis (PCA) as a preprocessing step for XCSR, and examine how it performs on complex multi-dimensional real-world data. The experiments show that this technique, in addition to significantly reducing the computational resources and time requirements of XCSR, maintains its high accuracy and even occasionally improves it. In addition to that, it reduces the required population size needed by XCSR.

Classification Based on LBP and SVM for Human Embryo Microscope Images

“…The feature extraction of embryo images has a very big influence on the classification, so finding a valid description of the images is a key step for classifying the embryos. Some commonly used feature extraction methods in pattern recognition, such as principal component analysis (PCA) [7] or linear discriminant analysis (LDA) [8], mostly depict an image from the overall point of view, which can well extract the global properties of the image but are sensitive to light and position. Moreover, the feature extraction method based on central moments [5,9] which has been used by some previous researches can quantify the key characteristics of embryo images in some extent, but it fails to consider the difference in the light intensity of embryo images.…”