Abstract: Principal Components Analysis (PCA) is a variable reduction technique helps to reduce a complex dataset to a lower dimensional subspace. This study is interested to investigate an approach for handling a problem occurred from considering a very large number of measured variables followed by a classification task. For such purpose, PCA has been used to extract and reduce of a very large number of variables that considered in the study. Then, a Linear Discriminant Analysis (LDA) which is commonly used for classification is constructed based on the reduced set of variables. The performance analysis of the constructed PCA+LDA was conducted and compared with the classical LDA Model using different size of sample (n) and different number of independent variables (p). The performance of PCA+LDA and classical LDA Model has been evaluated based on misclassification rate. The results demonstrated that PCA+LDA performed better than the classical LDA Model for small sample case. For large sample size case, PCA+LDA also performed better than the classical LDA especially when the measured independent variables is too large. The overall findings showed that the constructed PCA+LDA can be considered as a good approach for handling a very large number of measured variables and performing classification treatment.
Hashibah Hamid, Fatinah Zainon and Tan Pei Yong, 2016. Performance Analysis: An Integration of Principal Component Analysis and Linear Discriminant Analysis for a Very Large Number of Measured Variables. Research Journal of Applied Sciences, 11: 1422-1426.