Exploring Transparency in Decisions of Artificial Neural Networks for Regression

The interpretation and explanation of machine learning models are important tasks to ensure transparency and credibility. Various methods have been developed to achieve this goal, including Shap, Lime, and Explanatory Dashboards. The application of these explainability methods aims not only at interpreting the machine learning model but also at incorporating transparency into the implementation of a system that uses artificial intelligence algorithms. By leveraging part of a an initial manufactoring project, we implemented techniques to enhance the understanding of the model’s time prediction through explainability methods. This improves trust among the involved entities and contributes to the overall effectiveness of the system in various application contexts. Our objective is to apply those strategies to other projects with in industry 4.0 and manufactoring to increase the adoption of new and innovative understandable intelligent systems.