Abstract #

In today's industrial environment, the increased availability of real time data offers a great opportunity to perform datadriven decision making. And so, improving the manufacturing performance. Machine learning algorithms provide the ability to learn effectively from data. However, building an accurate machine learning model requires data scientits knowledge. Hence, the task of training an effective learning model becomes costly, time consuming, and laborious. Unfortunately, in several companies, industrial practitioners lack machine learning knowledge. In this context, we investigated to clarify the potential of Automated machine learning. AutoML is a set of tools that enable non-ML-expert to create automatically a successful ML model. Our study examines whether AutoML methods could achieve convincing results in various manufacturing applications such as quality control and predictive maintenance. We evaluated four AutoML tools: AMLBID, Autoweka, AutoSklearn, and TPOT on 7 industrial datasets. The experimental results have proved that some AutoML tools provided better performance than classical learning models with configuration performed by non-ML experts.

Citation #

@InProceedings{10.1007/978-3-031-23615-0_33,
author="Chaabi, Meryem
and Hamlich, Mohamed
and Garouani, Moncef",
title="Evaluation of AutoML Tools for Manufacturing Applications",
booktitle="Advances in Integrated Design and Production II",
year="2023",
publisher="Springer International Publishing",
address="Cham",
pages="323--330",
isbn="978-3-031-23615-0"
}