Developing Explainable Multimodal AI Models on the Life-Cycle of Rare Event Prediction

This hands-on lab explores the complete lifecycle of rare event prediction and analysis in industrial settings, addressing the $50 billion annual cost of unplanned downtime in Industry 4.0. Learn practical approaches to overcome the significant challenges in detecting and predicting extremely rare events that can cost industries like automotive over $2 million per hour in downtime. Gain experience with the entire process from data generation and preprocessing to model development and evaluation, including the development of process ontologies for user-level explanations. Work with limited publicly available datasets and get hands-on experience with a newly developed multimodal dataset specifically designed for rare event prediction. Through practical sessions, master techniques for generating high-quality datasets and applying them to develop effective rare event prediction models. The lab is valuable for anyone interested in developing AI models that leverage diverse multimodal data, whether for rare event prediction or other applications.