Cancer is one of the main global causes of death. Presently, Gene Expression Microarray (GEM) data has been used to assist an accurate and rapid detection of cancer and its subtypes. In various areas of biological research, the evaluation of gene expressions (GEs) is crucial for acquiring the relevant information. DNA microarray technology offers the ability to retrieve information from the expression levels of thousands of genes in a solitary experiment. Early detection of cancer and its subtypes can be directed in an ideal manner by the collection of relevant genes to increase the diagnostic accuracy. GEM data often generates tens of thousands of genes for each data sample. This results in lower sample size, high dimensionality issues and data complexity for detecting the cancer from GEM data. There is a need for computationally efficient and speedy methods to address these types of problems. So, an advanced Artificial Intelligence (AI) techniques such as Machine Learning (ML) and Deep Learning (DL) algorithms have been developed to deal with these issues. These models have achieved success in several disciplines including image, video, audio, and text processing. Similarly, ML and DL models address the challenges observed in (GEs) analysis for various cancer detection tasks to identify the most suitable biomarkers for the various cancer subtypes. This study provides a comprehensive analysis of the many ML and DL techniques designed to detect cancer and its subtypes by analyzing GEM data. Initially, multiple cancer detection and categorization models developed by numerous researchers using ML and DL algorithms are examined briefly. Then, a comparative research is undertaken to comprehend the shortcomings of these algorithms and to propose a new method for accurately detecting cancer and its subtypes