An Enhanced Clustering Method Based on Grid-Shaking

Clustering is an essential way to extract meaningful information from massive data without human intervention in the field of data mining. Clustering algorithms can be divided into four types: partitioning algorithms, hierarchical algorithms, grid-based algorithms, and locality-based algorithms. Each algorithm, however, has problems that are not easily solved. K-means, for example, suffer from setting up an initial centroid problem when distribution of data is not hyper-ellipsoid. Chain effect, outlier, and degree of density in data are problems occurring in other types of algorithms. To solve these problems, various kinds of algorithms were proposed. In this paper, we propose a novel grid-based clustering algorithm through building clusters in each cell and show how to solve the previously mentioned problems.