Application Areas, Use Cases, and Data Sets for Machine Learning and Artificial Intelligence in Production

Abstract

Over the last years, artificial intelligence (AI) and machine learning (ML) became key enablers to leverage data in production. Still, when it comes to the utilization and implementation of data-driven solutions for production, engineers are confronted with a variety of challenges: What are the most promising application areas, scenarios, use cases, and methods for their implementation? What openly available data sets for the training of ML and AI solutions do exist? In this paper, we motivate the challenges of applying AI and ML in production and introduce an extended taxonomy of application areas and use cases, resulting from a comprehensive literature review. In addition, we propose both a process model and a concept for an ML-Toolbox that are tailored to cope with the specific challenges of production. As a result, from an extensive study, we present and launch a comprehensive collection of currently more than 130 datasets that we make openly available online to serve as a continuously expandable reference for production data. We conclude by outlining three key research directions that are decisive for a widespread adoption of real-world ML. The contributions of this paper establish a foundational development framework that allows to identify suitable use cases, gain experience without having suitable in-house data at hand, improve existing data-driven solutions and promote applied research in this challenging field of ML in production.