The rapid expansion of information science has caused the issue of “the curse of dimensionality”, which will negatively affect the performance of the machine learning model. Feature selection is typically considered as a pre-processing mechanism to find an optimal subset of features from a given set of all features in the data mining process. In this article, a novel Hyper Learning Binary Dragonfly Algorithm (HLBDA) is proposed as a wrapper-based method to find an optimal subset of features for a given classification problem. HLBDA is an enhanced version of the Binary Dragonfly Algorithm (BDA) in which a hyper learning strategy is used to assist the algorithm to escape local optima and improve searching behavior. The proposed HLBDA is compared with eight algorithms in the literature. Several assessment indicators are employed to evaluate and compare the effectiveness of these methods over twenty-one datasets from the University of California Irvine (UCI) repository and Arizona State University. Also, the proposed method is applied to a coronavirus disease (COVID-19) dataset. The results demonstrate the superiority of HLBDA in increasing classification accuracy and reducing the number of selected features.2