A multi-objective feature selection algorithm using Sailfish Optimizer and Support VectorMachine

Feature selection involves removing the irrelevant features and selecting a distinct subset of features.Machine learning relies on feature selection, which has recently been examined as a multi-objectiveoptimization problem. Evolutionary techniques have solved feature selection problems because of theirhigh exploration capabilities. This paper proposes a feature selection approach using the SailfishOptimizer (SFO) algorithm and Support Vector Machine (SVM). Moreover, it tries to reduce theclassification error rate and the number of selected features in the dataset. The results of ۹ UCI datasetsshowed that the proposed algorithm was more accurate in most cases than five other meta-heuristicmethods (i.e., GA, GWO, EHO, MFO, and WOA). It provides very competitive performance whendealing with high-dimensional datasets. Additionally, as a no-free lunch theorem states that nometaheuristics-based feature selection technique can solve all optimization problems, the proposedmethod could be used for various feature selections on multiple datasets.

Keywords:

Feature Selection , Meta-heuristic , Optimization , Support Vector Machine