Classical Machine Learning Methods in Economics Research: Macro and Micro Level Examples

Abstract

Paper reviews the classical methods of machine learning (supervised and unsupervised learning),gives examples of the application of different methods and discusses approaches that will be useful for empiricaleconomics research (on data from Ukrainian firms, banks and official state statistics). The different sectors ofeconomics are investigated: the multiple linear regression is used on macrolevel for macro production functionof Ukraine specification; logistic regression is used in bank sector for credit risk management with the scoringmodel; k-means, hierarchic clustering and DBSCAN are used in regional level for regions of Ukraine groupingbased on competitiveness; principal component analysis is used for firm’s financial stability analysis. All modelsshowed adequate simulation results according to the quality criteria of the models. So, the possibility ofclassic machine learning methods application for investigations of the processes and objects on different levelsof economics (micro, mezzo and macro) is demonstrated in the article.