MicroGRID: An Accurate and Efficient Real-Time Stream Data Clustering with Noise

Data stream clustering aims to produce clusters from a data-stream in a real-time. Many of existing algorithms focus however on solving a single problem, leaving anomalous noise in data streams at the wayside. This paper describes the MicroGRID approach to cluster data from single data-streams to handle noisy data streams, accurately identifying and separating noise-affected data points from outlier points. In particular, MicroGRID utilises a combination of micro-cluster and grid-based prospectives, an approach that has not been attempted when clustering data-streams. The experimental results clearly show that MicroGRID significantly outperforms the baseline methods: MicroGRID is up 87% faster and up to 80% more accurate clustering outputs.