This dissertation extends discrete choice models and applies them to site selection by criminals. Spatial analysis of criminal incidents is an old and important application of spatial analysis. It uses past crime data to predict future crime locations and times. However, most of this analysis considers the aggregate behavior of criminals rather than individual spatial behavior. It does not consider the decision making processes of criminals as human initiated events susceptible to analysis using spatial choice models. Spatial analysis of crimes is different from other spatial choice analysis in important ways. The decision makers here are latent. It is impossible to estimate the characteristics of the population by sampled observations, which has been done in other spatial choice analysis.; In this work, the spatial choice modeling is extended to include the class of problems where the decision makers' preferences are derived indirectly through incident reports rather than directly through survey instruments. A methodology to analyze and predict the spatial behavior of the latent decision makers is provided by combining data mining techniques and the theory of discrete choice. For the spatial analysis of crimes, the decision makers of the criminal incidents are the unknown criminals. Their preferences can not be observed directly like the choice analysis in most other areas. Data mining techniques, like kernel estimation and feature selection are utilized to extract the preferences of the criminals. The spatial choice models are provided by adding the preference information to the combination of random utility structure and the hierarchical decision process of criminals.; In the criminal incident application, it is important to note that the decision makers will not have the same preferences over the study region. We use clustering techniques to discover and specify the preference heterogeneities of the latent decision makers. The prediction of the future crime locations are based on the preferences of all the specified decision makers. (Abstract shortened by UMI.)
展开▼
