A review study of microarray data classification with the application of dimension reduction / Sharifah Nadia Syed Hasan and Noor Wahida Jamil

Background. The growth of gene expression or microarray data, mainly in cancer disease, has become a game changer for feature selection techniques in handling complex data. Hence, the advancement of Deoxyribonucleic acid (DNA) microarray technology has made it feasible to measure the expression level of thousands of genes with the ability to diagnose early detection. This extensive study is conducted to review and analyse literature related to applying various dimensionality reduction approaches to predict microarray data. This study is aimed for the Data Science and Medical Sciences disciplines with the goal of extending future research and broader interdisciplinary collaboration efforts.

Methods. The systematic review of this study is based on the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines and reported in accordance with the PRISMA statement. Other than that, a systematic search is conducted using two search engines including, Scopus and Web of Science (WoS), from 2018 to 2022 by inputting the "feature extraction," "feature selection," "classification," and "microarray" as keywords. Based on the inclusion and exclusion criteria, the final articles available for review are 53 articles. Specifically, this study reports on the performance of feature selection approaches and the empirical comparisons of classification techniques used on the microarray dataset.

Results. According to the analysis, part of the included articles is mostly hybrid and novel approaches proposed for gene selection. Many novel and hybrid methods were developed to produce a good performance in terms of accuracy and computational efficiency. Moreover, the hybrid methods are proven effective in reducing dimensions and selecting relevant features. Besides, machine learning techniques are still the top interest among researchers for classification despite the emergence of deep learning approaches.