The Research of AutoML Methods in the Task of Wave Data Classification

This paper discusses automatic machine learning methods for solving the problem of objects’ classification. Automated machine learning is the process of automatically creating and optimizing machine learning models, capable of performing automatic feature selection, hyperparameter optimization, model selection, and model evaluation based on given quality metrics. In the task of classifying several signals, automating this process can be particularly useful, since this type of task can be complex to decompose and requires a lot of work to isolate meaningful features. In addition, AutoML has high applicability in the task of building neural networks, such as forward propagation neural networks, and optimizing hyperparameters such as the number of layers, the number of neurons. In this paper different approaches to AutoML have been analyzed. Methods based on MLJAR AutoML, AutoKeras and TPOT showed the best results.