Algorithms for detecting outliers via clustering and ranks
IEA/AIE'12 Proceedings of the 25th international conference on Industrial Engineering and Other Applications of Applied Intelligent Systems: advanced research in applied artificial intelligence
| Hi-index | 0.01 |
Density-based clustering and density-based outlier detection have been extensively studied in the data mining. However, Existing works address density-based clustering or density-based outlier detection solely. But for many scenarios, it is more meaningful to unify density-based clustering and outlier detection when both the clustering and outlier detection results are needed simultaneously. In this paper, a novel algorithm named DBCOD that unifies density-based clustering and outlier detection is proposed. In order to discover density-based clusters and assign to each outlier a degree of being an outlier, a novel concept called neighborhood-based local density factor (NLDF) is employed. The experimental results on different shape, large-scale, and high-dimensional databases demonstrate the effectiveness and efficiency of our method.