Addressing Gene Expression Profile Analysis Problems from a Factor Analysis Perspective

This study approaches gene expression profile analysis through factor analysis methodology. To resolve stability issues in conventional Independent Component Analysis (ICA), we propose a DNA microarray data integrated classifier based on selective ICA. The implementation involves analyzing reconstruction errors in gene expression levels, selecting ICs with minimal reconstruction errors for sample reconstruction, then training multiple Support Vector Machine (SVM) base classifiers concurrently using the reconstructed samples. Final classification is achieved through majority voting among high-accuracy base classifiers. Experimental validation across three benchmark datasets confirms the method's effectiveness.