Дослідження алгоритмів машинного навчання для побудови математичних моделей задач класифікації мультимодальних даних

Abstract

Currently, machine learning algorithms (ML) are increasingly integrated into everyday life. There are many areas of modern life where classification methods are already used. Methods taking into account previous predictions and errors that are calculated as a result of data integration to obtain forecasts for obtaining the classification result are investigated. A general overview of classification methods is conducted. Experiments on machine learning algorithms for multimodal data are performed. It is important to consider all the characteristics of metrics and features when using ML algorithms to predict multimodal data. The main advantages and disadvantages of Gradient Boosting, Random Forest, Logistic Regression and XGBoost algorithms are analyzed in the work.