Implementing machine learning algorithms to support workers in manufacturing remains a challenge. The main challenge is to find which algorithm configuration matches a certain type of use case. Currently, implementation projects focus on the specification of the algorithm configuration. Connection missing between algorithm configuration and machine learning solution.
This prevents the collaboration of domain experts, particularly when selecting the machine learning solutions (ML Solution).
We aim to determine the information that can make ML solutions and their functioning understandable to these domain experts. To achieve this, we propose a guiding process to direct the implementation of ML solutions. The process goes over the four main components of an ML solution: use case, data, infrastructure and algorithm. Each component's key information is encoded in a profile. These profiles are then used by an assessment function to predict the potential benefits and corresponding implementation effort that an ML solution requires before it is done. This information should make the use and selection of ML solution transparent and intuitive for domain-experts.
This project started on February 1st, 2018. Planned end date is June 30th, 2021.