Teaching Artificial Intelligence through Problem-Based Learning using Orange Data Mining: A Visual and Practical Approach

This paper presents a problem-based learning (PBL) strategy for teaching Artificial Intelligence (AI) through two practical case studies using Orange Data Mining-a visual and interactive platform that requires no coding. The approach engages undergraduate students with limited programming experience in real-world problems involving both regression and classification tasks. The first case involves predicting housing prices based on socio-economic and environmental variables. The second case focuses on fault diagnosis in Proton Exchange Membrane (PEM) Fuel Cell systems using real-time sensor data. In both scenarios, students design machine learning workflows, train models, and evaluate their performance using metrics such as precision, F1 scores and recall. Implemented in a third-year Bachelor AI course in ICAM Lille-Catholic Institute of Engineering and Technology, this approach enhanced students’ understanding of AI concepts, critical thinking, and modeling competence. Results from assessments and feedback suggest that combining visual tools with PBL creates an inclusive, engaging, and effective AI learning environment across domains including smart cities and sustainable energy.