Diverse Varieties of Cancer Forecasting System using Microarray Genes with the support of Improved Stacked Auto-Encoder

Abstract

Microarray gene expression data is one of the most commonly used gene expression datasets applied for cancer sample prediction. It consists of thousand expression levels of the genes in a single experiment. Cancer subtypes in microarray gene expression data are vague, indiscernible, imprecise and overlapping in nature which often lowers down the cancer prediction accuracy of the traditional classification models in general. Cancer prediction from gene expression data is an important and challenging area of research in the field of computational biology and bioinformatics. This research presents a deep learning approach to cancer detection, and to the identification of genes critical for the diagnosis of leukaemia,lung cancer and prostate cancer. The dimensionality of the data is handled using improved stacked auto encoder (ISAE) algorithm, whereby the features are classified using neural network.The proposed approach is enhanced by adding the regularization and reconstruction loss. Classification accuracy of ISAE is 97% that shows the effectiveness of the ISAE and this approach outperforms the existing state-of-art techniques.