With the increase in the dimensionality of data, feature selection has gained significant attention in recent times. Feature selection is a complex task due to the large search space involved. Swarm intelligence has recently been successfully used to address feature selection problems because of its global search ability. However, Beluga Whale Optimization (BWO), a swarm intelligence algorithm, is seldom used for feature selection due to its limited exploration ability at the end of iterations. Therefore, this paper proposes a new method called Memetic Beluga Whale Optimization (MBWO). By adding local perturbation and modifying search operators, MBWO can result in better performance in feature selection. Experiments comparing MBWO with other algorithms also show that the efficiency of the MBWO algorithm has significantly improved by these modifications.