應用地理剖繪分析臺北地區連續住宅竊盜的距離遞減型態與錨點的空間預測

Abstract

目標：「犯罪錨點」係指在連續犯罪事件中，犯罪者經常出現的地點，例如：居住地或工作地點等。由於台灣住宅竊盜案件呈現高發生率、低破獲率的趨勢，如何盡速預測與掌握連續住宅竊賊犯罪的空間行為與錨點，成為破案與預防竊案再次發生的關鍵。目前國內外文獻尚無完整且具體針對連續住宅竊盜錨點估計及空間型態的實證研究。因此，本研究目的在於分析連續住宅竊盜的犯罪錨點空間型態，找出罪犯錨點的可能分布區域，以縮小警方的罪犯搜尋範圍。 方法：地理剖繪（Geographical profiling）係以環境犯罪學為理論基礎，運用連續犯罪者的犯罪時間及空間型態等地緣因素，推估犯罪案件地點與罪犯錨點關連性的分析方法。本研究利用臺北地區在2006-2009年間，28個連續住宅竊盜犯罪者共326筆犯案地點和其居住地點的資料，應用地理剖繪的兩種分析方法：「犯罪旅程分析模式（Journey-to-crime analysis, 簡稱JTC）」和貝氏犯罪旅程分析模式（Bayesian journey-to-crime analysis, 簡稱BJTC），建立連續竊盜住宅犯罪的居住地點與犯案地點間的距離遞減關係，進而建立犯罪錨點的預測機率地圖。 結果：研究結果顯示以下五點：(1) 犯罪熱點的位置集中在台北市內湖區、南港區和文山區；新北市板橋區、中永和區、土城區和三重區，住竊犯平均犯案旅運距離為3.70公里。(2) 連續住竊犯的犯案空間行為確實有距離遞減現象，第一型離錨點0.5公里以上、第二型則離3公里以上始呈現距離遞減的趨勢。(3) BJTC考量各網格區域間的通勤關係和已破案起終點矩陣（Origin-Destination matrix），錨點推估機率地圖的空間分布型態呈現「非同心圓狀」。(4) 相較於BJTC，JTC的錨點準確度較高，JTC的錨點預測與實際值的誤差率小於1%，表示JTC確實能找出較正確的罪犯錨點。(5) 加入「警察局效果」等犯罪相關的建物環境參數，對於BJTC的推估效度具有顯著改善的效果。 結論：本研究應用地理剖繪方法，建立臺北地區連續住竊犯罪的錨點預測機率地圖，並區分不同距離遞減型態歸納連續犯的空間移動型態。本研究成果有助於提高警政單位犯罪偵察實務的效率，能主動出擊打擊犯罪，並進一步提供制訂治安政策的參考，以期達到降低住竊案的發生率。

Purpose: Anchor points of criminals are the zones where an offender travels routinely, such as work, school, home or recreational areas. The incidence rate of residential burglary is very high but detection rate is relatively low in Taiwan. In order to increase detection rate, it is important to understand spatial behaviors and anchor points of residential burglars. However, the relationships between crime locations of serial residential burglars and anchor points remain unclear. Therefore, the objective of the study is to identify spatial patterns and predict anchor points of serial residential burglars for targeting smaller zones of further criminal investigation. Methods: Geographical profiling is a criminal investigation methodology which is based on the theories of environmental criminology. Geographical characteristics and inter-relationships of serial criminal events are analyzed to determine the possible areas of offender residence. A total of 326 residential burglaries committed by 28 serial offenders were analyzed in this study. Journey-to-crime analysis (JTC) and Bayesian journey-to-crime analysis (BJTC) were used to determine the distance patterns between crime locations and residential areas of serial residential burglars and establish the probability maps for predicting criminal anchor points. Results: The significant findings indicated that (1) the hotspots of residential burglaries committed by serial offenders were identified and the average length of journey-to-crime trips was 3.7 kilometers; (2) these crime trips showed significant distance-decay relationships and could be divided into two types. Distance decay of Type I occurred at more than 0.5 kilometers while Type II occurred at more than three kilometers; (3) BJTC used Origin-Destination matrix to incorporate the commuting patterns of an offender, therefore, the probability maps of anchor point prediction did not show concentric circles; (4) JTC made more highly accurate prediction of anchor points than BJTC. The prediction error rate of JTC was less then 1 %; (5) the accuracy of BJTC could be improved significantly when the model was adjusted by crime-related built environmental factors, such as the locations of police stations. Conclusion: The study demonstrated that geographical profiling could be beneficial to predict the locations of anchor points of serial residential burglars in Taipei and differentiate distance-decay patterns of these crime trips. This study provides information valuable for the targeted delivery of crime investigation and prevention programs against residential burglaries.