中文摘要：

隨著犯罪地理學與環境犯罪學的理論軌跡匯合，以地理資訊系統製作犯罪地圖日臻普及與成熟，但如何將時間因素同時納入分析卻是棘手難題。為此，本研究使用掃描統計技術並配合地理資訊系統的製圖功能，針對臺北市2015至2017年住宅竊盜犯罪的空間、時間及時空分布現象進行分析。研究結果顯示，在空間分析方面，無論是使用圓形或橢圓形視窗掃描，臺北市住宅竊盜犯罪群聚熱區以中山區、萬華區、大同區及士林區較為顯著，且橢圓視窗掃描精確度略優於圓形視窗；在時間群聚部分，時間序列掃描顯示，歲末年終與跨年假期為住宅竊盜高發生時段；而在時空掃描方面，本研究使用「回顧性」與「前瞻性」掃描進行分析，在「回顧性」掃描部分，犯罪群聚的空間分布仍集中在中山區、萬華區及士林區等，但增加群聚時間的訊息後能更清楚判別群聚之特性；「時間趨勢的空間變化」與「前瞻性」掃描則發現在住宅竊盜犯罪冷區中，南港、內湖等部分地區有異常群聚現象，應特別予以關注，以達到「防微杜漸」的效果。
在了解時空中犯罪的分布特性後，本研究主要依據社會解組與新機會理論並參考相關文獻，選取各類社經人文變項進行區位分析，以釐清犯罪群聚形成之因素。在分析方法上，首先運用群聚共變項分析，以確認所選取之變項與本研究之群聚具有相關；其次，採用資料探勘技術與羅吉斯迴歸，篩選出犯罪群聚區位重要變項共計8項，結果顯示臺北市住宅竊盜犯罪冷區之特徵為高所得及房價較高地區有較低住宅竊盜犯罪；空屋率與單獨住戶比例較高地區，因監控力降低，易導致住宅竊盜犯罪；而在抑制犯罪的監控作為方面，犯罪熱區中有較多的警力配置，而路燈與監視器對於防竊的效果在本研究中並不明顯。
綜整前述之研究成果，本研究建議應針對不同熱區特性研擬具特色警務規劃，運用前瞻性掃描統計建構宅竊盜犯罪及時監測預警系統，並根據犯罪群聚最大概率與被害風險數據合理配置警政資源，以提升犯罪預防之成效。

英文摘要：

With the integration of the theory of criminal geography and environmental criminology, using geographic information systems to make crime maps has become increasingly popular and mature, but how to analyze time and space at the same time is a difficult problem. For this reason, this study used scan statistics and geographic information system to analyze the spatial, temporal and Spatiotemporal distribution of burglary in Taipei city from 2015 to 2017.The results show that, in terms of spatial analysis, whether it is using round or elliptical window scanning the hot spots of burglary are Zhongshan District, Wanhua District, Datong District and Shilin District, and the scanning accuracy of the elliptical window is slightly better than the circular window .In the time cluster, the time series scanning shows that the year-end and New Year holidays were hot time of burglary. And in terms of space-time scanning, this study was analyzed using both "retrospective" and "prospective" scans; In the "retrospective" scanning, the spatial distribution of burglary hotspots was still concentrated in the Zhongshan district, Wanhua District and Shilin District. However, the characteristics of clustering can be determined more clearly by increasing the information of time. "Spatial variation in the time trend" and " prospective " scanning found that in cold spots of burglary, some areas such as Nangang District, Neihu District have abnormal clustering phenomenon, which should be paid special attention to, so as to achieve the effect of "nip in the bud ".
After understanding the distribution characteristics of crime in time and space, this study selects various socioeconomic variables for ecological analysis based on criminology theory and relevant literature, so as to clarify the formation factors of crime clustering. Firstly, cluster covariant analysis was used to confirm the correlation between the selected variables and the clustering in this study. Secondly, data mining and logistic regression were used to screen out a total of 8 important ecological variations of crime cluster. The results show that the characteristics of burglary coldpots in Taipei city are in hight-income areas. The vacancy rate and the proportion of individual households are relatively high, resulting in reduced surveillance and easy to become burglary hotspots. There are more police forces in crime hotspots, while street lamps and monitors have no obvious effect on burglary prevention.
To sum up the above research results, this study suggests that special police planning should be developed for different crime hotspots, then prospective scan statistics should be used to construct a burglary timely monitoring and early warning system, and based on the maximum probability of crime and victim risk data to allocate police resources to enhance the effectiveness of crime prevention.

文章連結：

https://hdl.handle.net/11296/bkfp47

資料來源：

臺灣博碩士論文知識加值系統