In general, the initial population

In general, the initial population is very important for any MH method since it affects the convergence of an algorithm and the quality of the final solutions. According to the characteristics of the OBL strategy, it can be used to construct the initial population which accelerates the convergence rate.In this phase, the MFO starts by creating a population X, with N solutions and the dimension of each solutionis Dim as:X = L + (U − L) ⊙ rand(N, Dim), (10)where U is the upper bound, L is the lower bound of the search space, and ⊙ represents the commutative and associative product between two elements. To satisfy this step, the OMFODE method must convert the continuous values of the solutions into binary solutions, this is performed through using the following equation:For example, if we consider the values of xi are [0.93, 0.33, 0.61, 0.54, 0.28, 0.34, 0.86], then the output of Eq. (11) is represented as [1, 0, 1, 0, 0, 0, 1]. This representation means that the features in the dataset which are corresponding to ones are selected as relevant features and which are corresponding to 0’s will be ignored.After selecting the features, the fitness function is applied to evaluate the performance of these features, thefitness function used in our proposed algorithm is defined as in Eq. (12): where Errxi is the error of the classification process (i.e. K-NN classifier or any other classifier), |xi | defines the selected features number and Dim represents the total features’ number. ξ is a random value in the interval [0, 1],that is used to balance between the classifier’s accuracy and the number of the selected features.