For example,Бобцов

AUTOMATING THE CREATION OF MACHINE LEARNING MODELS FOR SOLVING TIME SERIES FORECASTING PROBLEMS

Annotation

It is proposed to use the AutoML approach to automate the creation of machine learning models predicting time series. Algorithms and technologies allowing to implement the automation of model creation are considered. A hybrid architecture of machine learning used in solving problems of automating the structural-parametric synthesis of models and optimizing hyperparameters, as well as in the automatic selection of indicators for assessing the quality of synthesized models, is chosen. The user interface for this system is implemented on the basis of the AutoGenNet platform, which embodies the No-Code development concept, which allows hiding the complexity of the processes of creating and training models from users, which reduces the entry threshold for working with the program. Using the AutoGenNet platform allows implementing a mechanism for generating software shells for operating trained models, as well as automating the processes of generating and training hybrid models, which simplifies and speeds up the process of solving time series forecasting problems using machine learning models. The results obtained can be scaled and used to create time series forecasting models in various applied problems.

Keywords

Постоянный URL

Articles in current issue

CONTENT © ITMO UNIVERSITY 1993-2025