An Self-Adaptive Data Stream Fuzzy Clustering Based on Granular-Ball Structure

Existing data stream clustering methods struggle to effectively address the issue of cluster boundary overlap caused by concept drift. Moreover, most methods require parameter settings, making them unsuitable for handling continuously evolving data streams. To address these challenges, we incorporate the concept of fuzziness into the granular-ball splitting process to mitigate the impact of cluster boundary overlap. Additionally, we leverage the adaptive splitting and merging process of granularballs to simulate micro-cluster changes in data streams. Finally, self-adaptive clustering is achieved based on the proximity rules between granular-balls. Experimental results on synthetic and real-world datasets demonstrate that, compared to existing methods, our approach achieves efficient and accurate clustering performance.