An Agent-Based Approach towards Crime Prediction: Evaluating the Potential of Environmental Factors and Human Dynamics to Model Offender Behavior

Crime reduction remains a critical challenge in modern societies. Thus, identifying regions in a city with an elevated risk for future crimes is of high interest to plan prevention strategies. Attempts have been made to apply various statistical and machine learning techniques to predict crime, while computational methods such as agent-based modeling (ABM) can account for the offender and its interactions with the environment in a new manner. In this thesis proposal, I describe an agent-based model to simulate crime, implementing offender mobility and offender decision-making informed by novel data and insights in human dynamics. The result will be a data-driven crime simulation for crime prediction, making it possible to study crime patterns and to reason about likely future high crime areas. This can then flow into a crime prediction tool for the police or into an experimental tool for policy makers and urban planners.