Abstract
Digital Twin (DT) is a growing concept in industries such as manufacturing, energy, smart cities, and built environments. To bridge the digital and physical worlds, DTs often rely on state-of-the-art Machine Learning (ML) models. However, many ML models, like Neural Networks, are considered black boxes, making their decision-making processes difficult to interpret. To address this, the explainable AI paradigm has introduced methods like SHAP and LIME. Despite their effectiveness, interpreting model outputs remains challenging. This paper explores the potential of using Tsetlin Machines in built environment DTs to explain the detection of structural properties. The Tsetlin Machine has demonstrated strong performance even with limited training data while offering both local and global interpretability. Its rule-based approach provides transparent decision-making, making it a promising alternative to traditional black-box models in DT applications.

This publication has 18 references indexed in Scilit: