DBSCAN Revisited MisClaim, UnFixability, and Approximationsigmod2015朱成名.pptxVIP

DBSCAN Revisited: Mis-Claim, Un-Fixability, and Approximation∗ Junhao Gan Yufei Tao Chinese University of Hong Kong New Territories, Hong Kong {jhgan, taoyf}@cse.cuhk.edu.hk Sigmod - 2015 CMZ 2015-10-27 2 Outline Definition and Background Existing approach Our approach Experiments Conclusion 3 Definition and Background DBSCAN : Density-Based Spatial Clustering of Application with Noise Grouping P into subsets—called clusters—such that any two clusters are separated by “sparse regions”.