The Classifier Models Usage for the Recruitment Process Forecasting for Applicants of Higher Education to Universities of Ukraine

The paper considers the task of classifying students entering the budget form of education using machine learning methods. Previous studies have been analyzed, which made it possible to develop the solving the problem strategy. Data for 2017–2020 Entrants has been collected thanks to open data. The data have been cleansed out of outliers, scaled using MinMaxScaler. Eliminated class imbalance using SMOTETomek methods. The specifics of various classifier models have been analyzed with the aim of better model selection and more detailed results analysis. Several classifier models have been trained, among them the utmost effective machine learning model was selected, which has reached out the Accuracy Level of 96% on the test sample, and 90% on the initial data. Hyperparameters for the final model have been reconciled. Its output values are calibrated for the purpose of more informative output of the model. The data processing algorithms and the model have been collected in the pipeline for convenient further use. The prototype of the program has been developed to demonstrate the model’s operability in conditions, which are close to the real ones. The anticipated input/output have been demonstrated while functioning, along with developing the scheme of the program’s operation in order to simplify its implementation into a real product. The conclusions have been made, thanks to which the context of the similar problems is being better understood, and their solutions can be used in a wider variety range of tasks.