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

Spatio-temporal association analysis has attracted attention in various fields, such as urban computing and crime analysis. The proliferation of positioning technology location-based services facilitated the expansion across spatio-temporal scales. However, existing methods inadequately consider scale differences among objects during analysis, leading to suboptimal precision results. To remedy this issue, we propose a multiscale object representation method, STO2Vec, for This method comprises two parts: graph construction embedding. For construction, introduce an adaptive hierarchical discretization distinguish varying scales local features. Then, merge embedding with that discrete units, establishing heterogeneous graph. embedding, enhance quality homogeneous data, use biased sampling unsupervised models capture strengths between objects. Empirical results using real-world open-source datasets show STO2Vec outperforms other models, improving accuracy by 16.25% on average diverse applications. Further case studies indicate effectively detects relationships range scenarios is applicable tasks moving behavior pattern mining trajectory semantic annotation.