Feature extraction in microarray gene expression profiling involves identifying relevant features for analysis and classification. The feature selection method outperforms gene selection methods in case of high-dimensional microarray data processing. This research presents dimensionality reduction methods, emphasizing their effectiveness in microarray feature selection. The proposed method uses three dimensionality reduction techniques, viz., Principal Component Analysis (PCA), Kernel PCA (KPCA), and Robust Kernel PCA (RKPCA). PCA and KPCA use covariance matrix to identify relationships between gene vectors with kernel function. On the other hand, RKPCA is an extension of KPCA that handles outliers and noise in microarray data. RKPCA uses kernel to map data into a higher-dimensional feature space, allowing the detection of non-linear gene relationships. RKPCA algorithm retains as much information bearing feature of spot as that of the original training data set. Hence, the microarray data classified using RKPCA algorithm has shown effective results when compared to PCA and KPCA. The accuracy of classification varies between 78% to 84% and 82% to 87% using PCA and KPCA respectively using 6 types of classifiers. Further, using RKPCA pre-processing method, classification accuracy has achieved commendable results of maximum 92.57% and 92.76% respectively using Adaptive Neuro Fuzzy Inference Systems (ANFIS).