Given a stream of real-time sensor data from a construction site, how would you design an efficient sliding-window anomaly detection algorithm?
