A Resilient BIRCH-Based Smart Framework for Real-Time IoT Data Clustering

Real-time data processing is essential in the evolving landscape of IoT applications, ensuring efficiency, reliability, and adaptability. However, conventional clustering algorithms often face difficulties in managing high-frequency, continuous IoT data streams due to limited adaptability and high computational overhead. To address these challenges, this study proposes a resilient adaptation of the BIRCH (Balanced Iterative Reducing and Clustering using Hierarchies) algorithm, tailored specifically for streaming IoT data. The enhanced approach dynamically recalculates clusters and determines the optimal number of clusters using the KneeLocator method. Unlike the original batch-oriented BIRCH, the modified version processes data incrementally, enabling continuous adaptation to changing data distributions. The proposed method was validated on benchmark IoT datasets and compared against K-Means, DBSCAN, standard BIRCH, and other state-of-the-art streaming-based clustering algorithms. Results consistently show that the modified BIRCH outperforms existing approaches in execution speed, memory efficiency, scalability, and clustering accuracy. In addition, the algorithm has been deployed within a web-based application featuring interactive visualization and anomaly detection, highlighting its practical relevance for smart city and industrial IoT scenarios. To promote reproducibility and future research, the complete framework and source code have been made publicly available.