Optimal Feature Selection Using Harris Hawk Optimization for Music Emotion Recognition

Feature selection is a fundamental and critical aspect of data preprocessing in the field of machine learning and data analysis. This selection not only reduces the dimensionality of the data but also improves the efficiency and effectiveness of various machine learning algorithms. Optimization techniques are commonly used in feature selection for classification problems to improve the performance of machine learning models by selecting the most relevant subset of features from a larger set of potential predictors. In this study, feature minimization is performed using the Harris hawk optimization (HHO) algorithm. HHO is a nature-inspired optimization algorithm that draws its inspiration from the hunting behavior of Harris hawks, a type of bird of prey. This algorithm was proposed in the field of computational intelligence to solve optimization problems. This paper aims at selecting the minimum number of features extracted from the Turkish music moods of 13 participants using this optimization. The feature number in the given dataset was reduced from 50 to 9, and the k-nearest neighbors algorithm (k-NN) achieved 85% classification accuracy. Therefore, it is expected that the HHO proposed in this study can play a very stable and effective role for feature selection problems.