Hybrid Binary Gray Wolf Optimization for finding Optimal Features in Classification Problems

Abstract

Listen

Translate article

Finding the essential symptoms(features) is highly demanded in the area of medical applications. Binary Gray wolf optimizer (BGWO) is one of the latest bio-inspired optimization techniques, which simulate the hunting process of gray wolves in nature. In this work, the binary gray wolf optimization (BGWO) is applied to select important feature subset for classification purposes and to attain maximum accuracy with minimum number of features using various classification algorithms and data sets. The classification error rate and the number of features are considered in the objective function. The wolf with low error rate and minimal features is considered as the best wolf and this kind of problem is a minimization problem. In BGWO, at each iteration the positions of three wolves (alpha wolf , beta wolf, delta wolf ) are identified. All the wolves move toward these three wolves to find out the target position. If the position of best solution (alpha wolf) is stuck in local minimum, the genetic algorithm (GA) is employed to get rid of it. The Hybrid binary gray wolf optimization with genetic algorithm (HBGWO) is used for classification problems in finding out the optimal feature subset with maximizing the classification accuracy while minimizing the number of selected features. The proposed HBGWO is used with the classification algorithms such as Naive Bayes, K-Nearest Neighbour and Decision Tree for different medical datasets. Results proved that the capability proposed HBGWO improves classification accuracy over BGWO.