Clustering based on grid and local density with priority-based expansion for multi-density data

Clustering based on grid and density for multi-density datasets plays a key role in data mining. In this work, a clustering method that consists of a grid ranking strategy based on local density and priority-based anchor expansion is proposed. In the proposed method, grid cells are ranked first according to local grid properties so the dataset is transformed into a ranked grid. An adjusted shifting grid is then introduced to calculate grid cell density. A cell expansion strategy that simulates the growth of bacterial colony is used to improve the completeness of each cluster. An adaptive technique is finally adopted to handle noisy cells to ensure accurate clustering. The accuracy, parameter sensitivity and computation cost of the proposed algorithm are analysed. The performance of the proposed algorithm is then compared to other clustering methods using four two-dimensional datasets, and the applicability of the proposed method to high-dimensional, large-scale dataset is discussed. Experimental results demonstrate that the proposed algorithm shows good performance in terms of accuracy, de-noising capability, robustness (parameters sensitivity) and computational efficiency. In addition, the results show that the proposed algorithm can handle effectively the problem of multi-density clustering.