A significant factor leading to the increased worldwide mortality rate is the tardy detection of cancer. Early detection of cancer can be achieved by the development of computer-aided diagnosis models utilizing machine learning (ML). Microarray data has a distinct and specialized use in conducting diagnostics. Despite the small sample size, microarray data contains several dimensions of a patient’s genetic information, including genes. Small sample sizes can be created by using microarray data as input for any machine learning model for classification without reducing dimensionality. Either a dimensionality reduction technique or a feature selection strategy must be utilized to boost the model’s performance. The microarray dataset is optimized and its dimensionality is reduced through the utilization of the Elephant Search Algorithm (ESA) in conjunction with four feature selection procedures: Maximum Relevance Minimum Redundancy (MRMR), Information Gain (IG), Correlation Feature Selection (CFS), and Recursive Feature Elimination (RFE). A support vector classifier has been utilized as a classifier. Three cancer microarray datasets have been analyzed. The efficacy of the hybrid approach has been assessed and contrasted using many metrics, including precision, accuracy, and recall. The MRMR feature selection technique attained an accuracy of 99.1%, which is higher than former feature selection techniques.