Domain Knowledge in CRISP-DM: An Application Case in Manufacturing

Abstract

To keep up with shifting technology trends and remain competitive, more manufacturing companies are investigating how to utilize data analytics to improve their processes. An issue these companies often face today is the need for more competence to perform advanced analytics projects within their departments. By using a human-in-the-loop approach and efficiently utilizing current domain knowledge in combination with data analytics, the higher success of implementation can be achieved. A common approach today to perform data analytics projects is to use the general Cross Industry Standard Process for Data Mining (CRISP-DM) methodology. This methodology does not consider the challenges specific to manufacturing and how to include domain expertise. This paper, therefore, suggests how the CRISP-DM methodology can be adapted to compensate for these issues. The adapted methodology is demonstrated in a case study for improving quality in the machining process by using interpretable machine learning models that can be used to assist experts when performing root cause analysis. This contributes to showing how to use domain experts’ knowledge better and how data analytics can be used in conjunction with domain-specific methods.