Paper Info
Title
Discovering Spatial Co-Location Patterns by Automatically Determining the Instance Neighbor.
Abstract
Spatial co-location pattern mining is an effective technique for identifying a group of features whose instances frequently located in geographical proximity. The distance between instances in space is commonly used to evaluate the proximity of them. If the distance between two instances is smaller than a distance threshold specified by users, they have a neighbor relationship. However, the neighbor relationships of instances determined in this way are ambiguous and it is difficult for users to set a suitable distance threshold. In addition, the neighbor relationships ignore the distribution of instances themselves, it is hard to deal with heterogeneous distribution density datasets. In this paper, we propose a new method based on the constrained Delaunay triangulation to determine neighborhoods of instances without distance thresholds for mining co-location patterns. First, applying the Delaunay triangulation to coarsely materialize the neighbor relationships of instances into an undirected connected graph. Then, we impose two strategies to constrain the Delaunay triangulation to prune undue edges. Finally, we develop a mining algorithm which can avoid the time-consuming generate-test candidate model which is widely used in previous algorithms. We perform experiments on both synthetic and real datasets to prove that our proposed method improves the accuracy and efficiency of mining results.
Year
DOI
Venue
2019
10.3233/FAIA190226
Frontiers in Artificial Intelligence and Applications
Keywords
Field
DocType
Spatial co-location pattern mining,constrained Delaunay triangulation,heterogeneous distribution density
Pattern recognition,Computer science,Artificial intelligence
Conference
Volume
ISSN
Citations 
320
0922-6389
0
PageRank 
References 
Authors
0.34
0
3
Name
Order
Citations
PageRank
Vanha Tran102.37
Lizhen Wang200.68
Hongmei Chen3255.39