Today, bioinformatics has become one of the important branches of interdisciplinary science that combines computer science and computational statistics to analyze data at the genomic, transcriptomic, and proteomic levels. One of the tasks performed in this field is the classification and examination of microarray data, which makes it possible to diagnose various diseases in the early stages of the disease by examining genome-level data. However, the classification of microarray data has many difficulties and complexities, and one of these difficulties is the existence of gene expression data sets with thousands of diverse genes. In order to diagnose a specific disease, it is necessary to identify and select limited and specific genes that play an effective role in causing that disease. Another challenge in this field is creating a model for diagnosing and predicting diseases using computer technologies, because most models require a large volume of data and samples for efficient diagnosis, and usually the number of samples related to a disease is limited and small, which makes modeling difficult. For this purpose, this article examines various gene selection and machine learning methods. A review of the results shows that most of these methods are based on classification and are efficient and highly accurate in building disease diagnosis models.