Data-centric optimization method to imbalanced datasets

Abstract

The imbalanced data sets is an important investigation in the data science and machine learning domain. This paper is focused on providing an effective algorithm for such cases. In these cases, the number of observations is not enough to construct a machine learning model with high values of the estimated magnitudes. For example, most of the observations are labeled as one class (majority class), and the rest as the other, commonly considered as the class of interest (minority class). We propose a methodology based on the properties of the data resampling approach to systematically enhance and optimize the training dataset. The paper examines numerical experiments applying the data centric optimization methodology and compares with previously obtained results by other authors.