Proposes a real-time adaptation of BIRCH for high-frequency IoT data streams. The Modified BIRCH dynamically recalculates clusters using the KneeLocator method, outperforming K-Means, DBSCAN and standard BIRCH in execution speed, memory efficiency and clustering accuracy across RT-IoT 2022, AirIoT and CIC IoTIDAD 2024 benchmarks.
IoT Clustering
Real-Time
BIRCH
Smart Cities
Anomaly Detection
DOI: 10.32604/cmes.2026.079203
Accepted: 23 March 2026