以GIS技術分析犯罪熱區：探討詐欺車手提款之時空環境因素與提領行為模式

Abstract

本研究聚焦於詐騙集團中負責資金提領的「ATM提款車手集團」，應用地理資訊系統（GIS）空間統計分析技術，以警政署165反詐騙專線彙整詐欺車手提領熱點資料為樣本，探討影響詐欺車手提領犯罪熱區之時空環境因素與提領行為模式，研究進一步分析環境、時間等變數與犯罪行為的空間關聯性，並以空間限制多變量集群分析進行分類，藉此預測與防範車手提領行為、提升查緝效率與策略規劃。研究結果顯示，詐欺車手的犯案行為並非隨機分布，而是深具地理策略意涵，提領車手傾向挑選兼具高空間可及性、交通便利、人潮密集且與低見警率風險的ATM據點作案，犯案時段則多集中於警力巡邏密度較薄弱的晚間與深夜，顯示其行為在空間與時間上皆具高度選擇性，再進一步的集群分析也指出，車手會根據不同地點的環境特性，如鄰近旅館、捷運站、交流道、銀行、超商、停車場或鄰近警察機關等因素，區隔化的提領模式，用以規避警方查緝、最大化其不法獲利，反映出其靈活調整的時空策略。透過標準差橢圓（Standard Deviational Ellipse）、空間自相關分析（Global& Local Moran’s I）、熱點分析（Getis-Ord Gi*）與空間限制多變量集群分析（Spatially Constrained Multivariate Cluster Analysis）等方法，本研究建構出詐欺車手犯案的時空環境模型，有效辨識出潛在高風險地點與時段，研究結果不僅驗證犯罪型態理論與日常活動理論對車手行為的解釋力，也凸顯詐欺車手「見風轉舵」的行動特性與對環境條件的高度敏感性。本研究成果有助於治安單位針對不同類型之提款熱區設計差異化的防制策略，提升巡邏與監控資源配置效率；同時亦提供犯罪地理學與環境犯罪學領域，針對詐欺行為提供更深層次的理解與理論基礎，未來相關研究與實務單位可進一步參考本研究所建立的分析框架，發展在地化的預警與防制模型，精準打擊詐欺犯罪。

This study focuses on the"ATM cash-out mule groups" within fraud syndicates and applies Geographic Information System (GIS) spatial statistical analysis techniques. Using hotspot data on fraud mules’ ATM withdrawals compiled by the National Police Agency’s 165 Anti-Fraud Hotline as the sample, the research investigates the spatiotemporal environmental factors influencing the formation of fraud withdrawal hotspots and the behavioral patterns of fraud mules. The study further examines the spatial correlations between environmental and temporal variables and criminal behavior, employing Spatially Constrained Multivariate Cluster Analysis to classify withdrawal patterns. The aim is to predict and prevent fraudulent withdrawals, thereby enhancing the efficiency of law enforcement and informing strategic planning.The results indicate that fraud mule activities are not randomly distributed but exhibit strong geographic strategic intent. Mules tend to choose ATM locations characterized by high spatial accessibility, convenient transportation, dense foot traffic, and low visibility to law enforcement. Their withdrawal activities are primarily concentrated during late-night hours, when police patrol presence is minimal, revealing a high degree of spatiotemporal selectivity. Cluster analysis further shows that mules tailor their withdrawal patterns based on local environmental features—such as proximity to hotels, metro stations, highway interchanges, banks, convenience stores, parking lots, or police facilities—demonstrating an adaptive strategy to evade detection and maximize illegal gains.Through the application of Standard Deviational Ellipse, Global and Local Moran’s I for spatial autocorrelation, Getis-Ord Gi* hotspot analysis, and Spatially Constrained Multivariate Cluster Analysis, this study constructs a spatiotemporal environmental model of fraud mule activity. The model effectively identifies high-risk locations and time periods. The findings support the explanatory power of Crime Pattern Theory and Routine Activity Theory in interpreting mule behavior and highlight the fraud mules' adaptive tactics and sensitivity to environmental conditions. This research contributes practical insights for law enforcement agencies to design differentiated prevention strategies for various withdrawal hotspots and optimize patrol and surveillance resource allocation. Moreover, it offers a deeper understanding and theoretical foundation for the fields of crime geography and environmental criminology. Future research and practitioners may adopt the analytical framework established in this study to develop localized early warning and prevention models for more precise counter-fraud interventions.