ADCN: An anisotropic density-based clustering algorithm for discovering spatial point patterns with noise

In this work we introduce an anisotropic density-based clustering algorithm. It outperforms DBSCAN and OPTICS for the detection of anisotropic spatial point patterns and performs equally well in cases that do not explicitly benefit from an anisotropic perspective. ADCN has the same time complexity as DBSCAN and OPTICS, namely O(n log n) when using a spatial index, O(n2) otherwise. STKO@Geography UCSB INTRODUCTION Cluster analysis is a key component of modern knowledge discovery. A wide range of clustering algorithms, such as DBSCAN, OPTICS, Kmeans, and Mean Shift, have been proposed and implemented over the last decades. Many clustering algorithms assume isotropic secondorder effects among spatial objects thereby implying that the magnitude of similarity and interaction between two objects mostly depends on their distance. However, the genesis of many geographic phenomena demonstrates clear anisotropic spatial processes. Isotropic clustering algorithms such as DBSCAN have difficulties dealing with the resulting point patterns and either fail to eliminate noise or do so at the expense of introducing many small clusters. To address this problem, we propose an anisotropic density-based clustering algorithm. Fig.3 Ground truth and best clustering result comparison for 6 synthesis cases and 4 real world cases. Fig.1 Illustration for ADCN-Eps.