Master of Science (MS)
Geography and Anthropology
A series of hotspot mapping theories and methods have been proposed to predict where and when a crime will happen. Each method has its strengths and weaknesses. In addition, the predictive accuracy of each hotspot method varies depending on the study area, crime type, parameter settings of each method, etc. The predictive accuracy of hotspot methods can be quantified by three measures, which include the hit rate, the predictive accuracy index (PAI), and the recapture rate index (RRI). This thesis research applied eight hotspot mapping techniques from the crime analysis field to predict crime hotspot patterns. In addition, these hotspot methods were compared and evaluated in order to possibly find a single best method that outperforms all other methods based on the three predictive accuracy measures. Identifying the single best method is carried out for all Part1 Crimes combined and individually, for five of the nine Part 1 Crime. In addition to the spatial analysis, a spatial–temporal analysis of the same crime dataset was conducted to investigate the distribution of crime clusters from both the space and time dimensions. The reported crime data analyzed in this study are from the city of Houston, TX, from January 2011 to December 2012. The results show that the predictive accuracy is affected by both the hotspot mapping method and the crime type, although the crime type has a more moderate effect. Considering the use of the three predictive accuracy measures, the kernel density estimation could be identified as the method which could most accurately predict the overall Part1 Crimes for the city of Houston. The nearest neighbor hierarchical clustering and kernel density estimation could be identified as the methods which are best at predicting each of the five crime types examined based on PAI and RRI, respectively. Also, spatial-temporal analysis indicates that more crimes occurred during September to December, 2011 around the center and in the southwestern part of the city of Houston, TX.
Document Availability at the Time of Submission
Release the entire work immediately for access worldwide.
Fan, Shuzhan, "The spatial-temporal prediction of various crime types in Houston, TX based on hot-spot techniques" (2014). LSU Master's Theses. 219.