Boosting density-based clustering algorithm by mean approximation on grids

Many applications require clustering of large amount of data. Most clustering algorithms, however, do not work and efficiently when facing such kind of dataset. This paper presents an approach to boost density-based clustering algorithms by mean grids. We show that the method of adjusted mean approximation on the grid is not only a powerful tool to relieve the burden of heavy computation and memory usage, but also a close proximity of the original algorithm. An adjusted mean approximation on grids using density-based clustering algorithm (AMADC) is constructed which exploits more advantages from the grid partition mechanism. Results of experiments also demonstrate promising performance of this approach.