Discovering Spatial Co-Location Patterns Under Considering the Distribution of Spatial Features

As science and technology advance, massive data, such as spatial data in smart city application, have been collected. It has gradually become an important problem to mine prevalent co-location patterns from massive spatial data. Spatial co-location pattern mining is a very important research branch in the field of spatial data mining. A spatial co-location pattern is a set of spatial features which instances have frequently neighbor relationships each other in space. The traditional co-location mining method usually gives a single neighbor threshold to measure the neighbor relationship between instances, and the spatial distribution of instances of different spatial feature is not to be consistent. To solve the above problem, as explained by Tobler's first law, with the distance increases, the contribution of instances to the importance of patterns decreases, so the distance weight of instances is proposed. The influential threshold is defined from the distribution of feature instances. A spatial co-location pattern mining algorithm based on kernel density estimation model with influential threshold is proposed. Experiments are carried out in synthetic and real data sets. The results show that the algorithm improves effectiveness and efficiency compared with the traditional methods.