HFMOEA: a hybrid framework for multi-objective feature selection

Abstract In this data-driven era, where a large number of attributes are often publicly available, redundancy becomes major problem, which leads to storage and computational resource requirement. Feature selection is method for reducing the dimensionality data by removing such redundant or misleading attributes. This optimal feature subsets that can be used further computation like classification data. Learning algorithms, when fitted on reduced dimensions, perform more efficiently storing also easier. However, there exists trade-off between features selected accuracy obtained requirement different tasks may vary. Thus, in paper, hybrid filter multi-objective evolutionary algorithm (HFMOEA) has been proposed based nondominated sorting genetic (NSGA-II) coupled with filter-based ranking methods population initialization obtain an solution set problem. The two competing objectives minimization maximization accuracy. help faster convergence NSGA-II PF. HFMOEA evaluated 18 UCI datasets 2 deep sets (features extracted from image using learning models) justify viability approach respect state-of-the-art. relevant codes available at https://github.com/Rohit-Kundu/HFMOEA.