A Streaming IRFCM Framework for Real-Time Anomaly Detection in IoT Sensor Networks

The proliferation of Internet of Things (IoT) devices has led to unprecedented streams of high-dimensional, noisy sensor data that require advanced clustering and anomaly detection techniques. This study proposes a novel Streaming Integrated Robust Fuzzy Clustering Module (S-IRFCM), extending the IRFCM framework to enable adaptive, real-time soft clustering and anomaly detection in large-scale IoT sensor networks. S-IRFCM integrates online dimensionality reduction, incremental centroid updating, and dynamic outlier trimming for robust, efficient, and interpretable clustering. Experimental evaluation demonstrates superior robustness in noisy, drifting environments where standard streaming baselines fail, alongside orders-of-magnitude faster processing than batch fuzzy methods. These results confirm that S-IRFCM achieves scalable, stable clustering in dynamic and noisy big data environments, providing a new paradigm for IoT-enabled analytics applications.