Discovering regions of anomalous spatial co-locations

STO2Vec: A Multiscale Spatio-Temporal Object Representation Method for Association Analysis

“…The strength of association among these objects is reflected by the cosine similarity of their embedding vectors: the stronger the association is, the higher the cosine similarity. The strength of spatio-temporal association can be reflected by the spatiotemporal co-occurrence frequency [7], which is similar to the learning of word vector representation obtained by word co-occurrence laws in natural language processing. Therefore, we introduce the Skip-gram [44] model in conjunction with the heterogeneous graph constructed in Section 4.3.…”

“…Associations are universal [3], making association analysis applicable to various fields with different relationship types. Sharma et al [2] grouped spatio-temporal associations into three types based on whether a temporal sequence was considered: sequential (e.g., analyzing event-oriented spatio-temporal association in video surveillance [4]), cascading (e.g., studying relationships between events, locations, and criminal activities in criminal geography [5]), and co-occurrences (e.g., similar associations between trajectories [6], co-location patterns between geographic entities [7], semantic annotation of trajectories [8][9][10][11][12][13][14][15][16][17][18][19][20][21][22], and location embedding [23][24][25][26][27][28][29][30][31][32][33], etc.). By comparing their frequency of co-occurrence, spatio-temporal co-occurrence-based association analysis can reveal implicit associations between entities.…”

“…Thus, the adaptive discretization problem of multiscale spatio-temporal objects can be transformed into an optimization problem with constraints of the form shown in Equation (7).…”

Mining significant local spatial association rules for multi-category point data

“…To address the problem that mining global spatial colocation patterns can not effectively identify significant spatial colocation patterns in local regions, scholars have carried out a large number of studies on mining local spatial colocation patterns [17] . In early studies on mining local spatial colocation patterns, the research area was divided into subunits by partitioning, and the local spatial colocation pattern was identified in each subunit by global spatial colocation pattern mining algorithm.…”