An Enhanced Version of MDDB-GC Algorithm: Multi-Density DBSCAN Based on Grid and Contribution for Data Stream

Abstract

With the continuous enrichment of big data technology application scenarios, the clustering analysis of a data stream has become a research hotspot. However, the existing data stream clustering algorithms usually have some defects, such as inability to cluster arbitrary shapes, difficulty determining some important parameters, and “static” clustering. In this study, a novel algorithm MDDSDB-GC is innovated. It selected MDDB-GC as the original algorithm that cannot deal with a data stream. In MDDSDB-GC, the calculation methods of contribution, grid density, and migration factor are effectively improved, and other parts are adjusted accordingly. The experiments show that MDDSDB-GC retains the advantage of MDDB-GC and obtains the ability to cluster an analysis for a data stream. At the same time, it effectively overcomes the above conventional defects, and its overall performance is better.