Gene Expression Data Classification with Kernel independent Component Analysis
Abdallah Bashir Musa · 2014
The challenge of classifying the characteristics of gene expression data is that the size of the training data is significantly lower than the number of features. Logistic regression (LR) is standard statistical method that broadly used in medical, epidemiology and bioinformatics communities for classification task; however, in such situation of gene expression data, LR does not work efficiently due to multi- collinearly and over- fitting problems, therefore, modifying of LR to analysis the microarray data is required. For solving those problems, reduction dimension is usually used. Recently, kernel approaches have proven to be good for classification such type of data. Kernel independent component analysis (KICA) is the nonlinear form of independent component analysis (ICA). In this paper, LR is applied to classify the features that selected by KICA. To evaluate the classification performance of this technique, this method has compared to kernel principle component analysis (KPCA) and independent component analysis (ICA). Numerous performance metrics such as accuracy, sensitivity, specificity, precision, F-score, the area under receiver operating characteristic curve (AUC) and the receiver operating characteristic (ROC) analysis are used.