BayesianOptFs: Bayesian Optimization Based Novel Feature Selection Method for Classification Models

Nowadays, machine learning models are frequently employed in high-dimensional datasets. At this stage, it is known that feature selection methods make significant contributions to the performance of machine learning models. In this study, a unique feature selection method based on Bayesian optimization was developed to contribute to the feature selection literature, and the method was tested using nine classification methods on two datasets. According to the experimental results, it was concluded that the proposed method provides a substantial decrease in the number of features without causing significant drops in performance scores. The aim is to contribute to the literature by sharing the source codes of the proposed method on open-source platforms. The fact that the Bayesian optimization-based method developed in the study has not been developed before is considered the novelty of the study. Additionally, sharing the method in a reusable manner on open-source platforms is anticipated to increase the prevalence of the method.