A Situational Awareness‐centric Predictive System for Anomaly Detection
Abhishek Djeachandrane, Said Hoceini, Serge Delmas, Abdelhamid Mellouk · 2025
This chapter describes a novel situational awareness-centric predictive system. Machine learning methods are known as an artificial intelligence driven by data. Access to the data is the core problem with this type of system. The chapter focuses on four problems that should be studied and solved as a whole, individually, or partially: temporal anomaly localization, i.e. “credit assignment”; trial-and-error learning, i.e. “no teachable by example”; concept drift; and model interpretability. To answer mission-critical properties, two approaches will be presented. The first one is the need for an interpretable model, and the second one is the need for an adaptive model. To answer mission-critical performance needs in terms of accuracy, two approaches will be presented. The first one is known as a modular approach, and the second is known as the customized learning scheme approach.