A Robust Adaptive Hierarchical Learning Crow Search Algorithm for Feature Selection

Feature selection is a multi-objective problem, which can eliminate irrelevant and redundant features improve the accuracy of classification at same time. great challenge to balance conflict between two goals feature ratio. The evolutionary algorithm has been proved be suitable for selection. Recently, new meta-heuristic named crow search applied problem This advantages few parameters achieved good results. However, due lack diversity in late iterations, falls into local optimal problems. To solve this we propose adaptive hierarchical learning (AHL-CSA). Firstly, an technique was used divide population several layers, with each layer from top particles topmost other. strategy encourages more exploration by lower individuals exploitation higher individuals, thus improving population. In addition, order make full use information level reduce impact optimization on overall performance algorithm, introduce sharing mechanism help adjust direction convergence algorithm. Finally, different difference operators are update positions levels. further improved using operators. method tested 18 standard UCI datasets compared eight other representative algorithms. comparison experimental results shows that proposed superior competitive Furthermore, Wilcoxon rank-sum test verify validity