BeSTAD: Behavior-Aware Spatio-Temporal Anomaly Detection for Human Mobility Data

TL;DR

BeSTAD addresses the challenge of identifying individual-level behavioral anomalies in large-scale mobility data by learning personalized normal patterns and performing cross-period comparisons with semantic alignment. It combines multi-scale spatial semantics from OpenStreetMap with temporal features in an LSTM-based clustering framework built on VAMBC, producing semantically enriched representations and behavior profiles. Anomalies are detected by aligning cluster semantics across periods and aggregating six personalized behavioral-change scores into a single $anomaly$ score, enabling interpretable, per-person insights. On the NUMOSIM benchmark, BeSTAD substantially outperforms a context-aware trajectory baseline in AUROC and demonstrates the benefit of fusing spatial semantics with individualized behavior modeling for practical mobility surveillance.

Abstract

Traditional anomaly detection in human mobility has primarily focused on trajectory-level analysis, identifying statistical outliers or spatiotemporal inconsistencies across aggregated movement traces. However, detecting individual-level anomalies, i.e., unusual deviations in a person's mobility behavior relative to their own historical patterns, within datasets encompassing large populations remains a significant challenge. In this paper, we present BeSTAD (Behavior-aware Spatio-Temporal Anomaly Detection for Human Mobility Data), an unsupervised framework that captures individualized behavioral signatures across large populations and uncovers fine-grained anomalies by jointly modeling spatial context and temporal dynamics. BeSTAD learns semantically enriched mobility representations that integrate location meaning and temporal patterns, enabling the detection of subtle deviations in individual movement behavior. BeSTAD further employs a behavior-cluster-aware modeling mechanism that builds personalized behavioral profiles from normal activity and identifies anomalies through cross-period behavioral comparison with consistent semantic alignment. Building on prior work in mobility behavior clustering, this approach enables not only the detection of behavioral shifts and deviations from established routines but also the identification of individuals exhibiting such changes within large-scale mobility datasets. By learning individual behaviors directly from unlabeled data, BeSTAD advances anomaly detection toward personalized and interpretable mobility analysis.

Figures (1)

• Figure 1: Overview of BeSTAD. The framework integrates spatial and temporal features to detect individual-level mobility anomaly detection through cross-period behavioral modeling.