中文摘要：

根據研究，犯罪並不是均勻的散佈在各個角落，而是有「聚集」的現象，即相同的地點一再發生犯罪事件。因此要達到預防犯罪的最大效果，必須分析出犯罪發生的聚集地點，也就是犯罪熱點（Hot Spots）。本研究以民國104至107年台北市住宅竊盜為例，希望透過犯罪熱點的分析，找出熱點聚集區域，並藉由相關性分析挖掘高度影響都市犯罪的相關因素，為當地預防犯罪提供依據。
本研究基於官方數據，首先利用GIS軟體對台北市民國104至107年間住宅竊盜數據進行處理後，將其製作成可視化的犯罪地圖，初步了解犯罪點位在地圖上分布的情形。而後藉由「Global Moran’s I」與「平均最近鄰分析」檢驗出住宅竊盜犯罪有群聚的現象，接著為進一步詳細分析，本研究將住宅竊盜資料分為「點資料」與「面資料」進行分析處理。
在「點資料」分析中，利用核密度推估方法計算，描述犯罪空間風險的連續分佈，將台北市的住宅竊盜熱點可視化，但核密度推估法無法將犯罪冷區顯示在地圖，故進而在「面資料」分析中利用Anselin’s LISA和Getis-Ord’s Gi方法加以分析，兩種方法在熱點分布上的結果相互呼應。
但透過上述的研究方法皆只能得到住宅竊盜是否有群聚的現象，以及犯罪熱點位於台北市的何處，而本研究為進一步探討何種因素會影響台北市的犯罪數量，因此除了以106年住宅竊盜與透過文獻回顧所選的土地面積、教育程度、警力、人口、娛樂場所，五個類別因子進行相關性分析外，也將106年自行車竊盜數、106年汽車竊盜數，共同加入討論。透過相關性分析後，選取相關係數高的因子，將三種竊盜犯罪數做為依變數，並將因子做為自變數，操作逐步迴歸分析，可以得到以下結論：
（一）商業區面積、工業區面積、每萬人員警數、合法八大行業數與住宅竊盜、自行車竊盜和汽車竊盜犯罪數量之間的相關度均不高。
（二）合法電子遊戲場數雖然與住宅竊盜有高度負相關，其值為-0.904；也與汽車竊盜有高度正相關，其值為0.990。但因為顯著性都未通過檢驗，因此不具有統計意義。
（三）可以用台北市各區住宅面積較好地預測出行政區內的住宅竊盜數量；可以用各區的博士數量較好地預測出行政區內的自行車竊盜數量；可以用各區的最高學歷國小畢業人數、性別比較好地預測出行政區內的汽車竊盜數量。

英文摘要：

According to study, the distribution of crime at all corners is not even but “clustering”, which means there would be same crime in the same place. Therefore, to reach the maximum effect of crime prevention, the clustering place of the occurrence of crime, the Hot Spots, must be found out. In the study, the burglaries in Taipei from 104 to 107 years of the Republic of China are example in the study. Hope that the clustering sites could be found out through the analysis of hot spots and dig out the factors that would highly affect the metropolis crime with correlation analysis, so as to provide evidence for the local crime prevention.
Based on official data, the study have processing of the burglary data in Taipei from rom 104 to 107 years of the Republic of China with GIS software. Then the data is made as a visual crime map to have a primary understanding of the distribution of the hot spots on the map. Next, the clustering of burglary is verified by the Global Moran’s I and Average Nearest Neighbor Analysis and there is further specific analysis. In the study, the burglary document are divided into Point date and Polygon date to have analysis and processing.
In the analysis of Point date, the sustainable distribution of crime space risk is calculated with Kernel Density Estimation, so that the hot spots of burglary in Taipei could be visualized. However, the Kernel Density Estimation could not show the cold area of crime on the map, so Anselin’s LISA and Getis-Ord’s Gi are used in the analysis of Polygon date. The results of the two methods correspond to each other.
However, the study methods mentioned above could only confirm whether there is clustering atmosphere of burglary and the specific area of the hot spots in Taipei; while the study will have further discussion on what factors would affect the number of crimes in Taipei. Therefore, in addition to the correlation analysis of the five categories of factors, including land area, education level, police force, population and places of entertainment selected through the literature review and the burglary in 106 years of the Republic of China, the number of bicycle theft and vehicle theft in 106 years of the Republic of China are also added to the discussion. Factors with high correlation coefficient are selected through the correlation analysis. Three kinds of theft crimes are made as dependent variables and the factors are independent variables to have stepwise regression analysis. The following conclusions are gained the study:
(I) Commercial area, industrial area, number of police per 10000 people, number of specialized service occupation are not highly correlated with the number of burglaries, bicycle theft and vehicle theft crimes.
(II) There is a high negative correlation between the number of legitimate video games and burglary with the value of -0.904 and there is also high positive correlation between the number of vehicle theft and burglary with the value of 0.990, but the significance could not pass the examination, so there is no
statistical significance.
(III) Residential area of Taipei could be used for better prediction of the number of burglaries within the administrative area. Number of PhDs in each district could be used for better prediction of the number of bicycle theft. The number and gender of graduates from primary and secondary schools with the highest education background could be used for better prediction of the number of vehicle theft within the administrative area.

文章連結：

https://hdl.handle.net/11296/r4yp73

資料來源：

臺灣博碩士論文知識加值系統