APPLICATION OF THE MULTIDIMENSIONAL KERNEL DENSITY ESTIMATION METHOD IN MACHINE LEARNING TASKS WITH IMBALANCED DATA

Popukaylo Vladimir Sergeevich, Shmelyova Anastasiya Vladimirovna

The article addresses the problem of using imbalanced data in multi-class classification tasks. It briefly examines the main existing approaches and proposes the application of the multidimensional kernel density estimation method to balance classes. The algorithm for applying this method is described, and an experiment is conducted using synthetic data. The results are compared with existing algorithms such as random oversampling of the small class, ADASYN, SMOTE, ASMO, SVMSMOTE. The article shows the possibility of using the multidimensional kernel density estimation method in principle to improve the quality of machine learning algorithms in conditions of imbalanced data

Keywords:Machine Learning, classification task, tabular data processing, imbalanced data, multidimensional kernel density estimation method