Degree
Doctor of Philosophy (PhD)
Department
Geography and Anthropology
Document Type
Dissertation
Abstract
This dissertation sets out to develop a new method that further improves the spatiotemporal kernel density estimation (STKDE) method by taking into account of the commonly known cyclical trend in crime occurrence, therefore termed the “STKDE-ct method”. This study utilizes a data-driven spectral analysis approach, periodogram, to detect the absence or presence of a periodic pattern in crime. The result shows that robbery and aggravated assault have a 6-month period of statistical significance in Baton Rouge from Jan. 2010 to May 2018. This identified period is incorporated into the temporal dimension of STKDE-ct method. The temporal kernel function of the traditional STKDE only considers a monotonic trend, whereas the STKDE-ct method has a new temporal kernel capturing both the linear and cyclical trends of crime occurrence.
The robbery in Baton Rouge city is selected as an example to implement the STKDE-ct method and compares its performance to those of the STKDE and KDE methods. Forecast Accuracy Index (FAI) and Forecast Precision Index (FPI) are used to measure the performances of the three methods in accuracy and precision, respectively. Comparing the hotspot maps of the three methods confirms that the predictive accuracy and precision have some trade-off effects as suggested by the literature. Based on the results from the crime type and study area in this dissertation, the STKDE-ct method outperforms the STKDE method in both predictive accuracy and precision on average, the STKDE-ct method also outperforms the KDE method in predictive accuracy on average, while the STKDE-ct generates results of less precision than those by the KDE on average. When looking to the performance across specific weeks, there are a significant number of weeks that the STKDE-ct has the best performance in both predictive accuracy and precision. The scenario with the best predictive accuracy tends to provide a more consolidated hotspot pattern (a few relatively large hotspots), while the scenario with the best predictive precision tends to generate a more scattered hotspot pattern (a large number of small hotspots). Among the three methods implemented on the robbery crime in Baton Rouge, the scenario with the best predictive accuracy by the STKDE-ct is considered more feasible in implementation of hotspot policing in practice.
Date
8-23-2021
Recommended Citation
Han, Ya, "Predictive Crime Hotspot Analysis with the Spatiotemporal Kernel Density Estimation Method Accounting for a Cyclical Trend: A Case Study in Baton Rouge" (2021). LSU Doctoral Dissertations. 5646.
https://repository.lsu.edu/gradschool_dissertations/5646
Committee Chair
Wang, Fahui
DOI
10.31390/gradschool_dissertations.5646