Resumen: Esta nota de política resume los resultados preliminares del proyecto “Diseño y validación de modelos de analítica predictiva de fenómenos de seguridad y convivencia para la toma de decisiones en Bogotá” (véase Acerca del Estudio) y algunas de sus aplicaciones: (1) despliegue óptimo de la policía por cuadrante y turno, (2) ubicación y monitoreo óptimo de las cámaras de seguridad de la ciudad y (3) ubicación óptima de los Centros de Atencion Inmediata CAI y/o estaciones de policía.

Resumen: Not all types of crime have the same priority in the agendas of policymakers since society tends to be more reluctant to more violent and costly crimes such as homicide. However, relative to other types of crime, homicides are statistically more challenging due to its sparsity and low frequency. For instance, over the last five years the average number of homicides across the city of Bogota has been roughly a thousand events per year, compared to the more than one hundred thousand robberies reported in the same period. Nevertheless, more than 80% of the homicides in the city occur during street fights suggesting a strong spatial and temporal correlation between these two types of crime. With this in mind, we used a manifold learning approach that capitalizes on a rich dataset of street fights to discover a criminal manifold that we use to penalize a KDE model of homicides where sparsity and low frequency is an issue. To implement this we follow a Kernel Warping methodology (Zhou & Matteson, 2015). The methodology reduces the relevant space for homicide prediction to regions of the city where homicides or street fights have occurred, giving more weight to the homicide episodes. We also introduce a temporal decay component to place a larger importance to recent events. The proposed model outperforms a standard KDE trained with homicide data, a KDE trained in both homicide and street fights data for homicide prediction, and a standard self-exciting point process on homicide data: flagging just the 5% of the area of the city with the highest estimated density, the Kernel Warping model correctly identifies between 30% and 35% of the homicides in the test set.

Resumen: We propose a novel conditional GANs architecture for crime (robberies) prediction in Bogotá, capital city of Colombia. The model uses several layers of ConvLSTM neural nets in both the generative and the discriminatory networks. We further condition on past crime intensity maps, weekdays, and holidays. The trained network is able to capture spatiotemporal patterns and outperforms state-of-the-art predictive models such as spatiotemporal Poisson point process, as well as other models trained with the same dataset. Model's accuracy reaches an area under the Hit Rate - Percentage Area Covered by Hotspots curve of 0.86. However, our predictions suggest that there is a potential bias with heterogeneous effects on vulnerable populations. We address the fairness consequence of this model in low income vs. high income residents by estimating a calibration test conditional to these protected variables. Finally, we introduce a fairness - accuracy balancing technique that quantifies the tradeoffs between accuracy and fairness in this type of models. This technique notably reduces bias with a marginal effect on accuracy.

Resumen: I use residential burglary data from Bogota, Colombia, to fit an agent-based model following truncated Levy flights (Pan et al., 2018) elucidating criminal rational behavior and validating repeat/near-repeat victimization and broken windows effects. The estimated parameters suggest that if an average house or its neighbors have never been attacked, and it is suddenly burglarized, the probability of a new attack the next day increases, due to the crime event, in 79 percentage points. Moreover, the following day its neighbors will also face an increment in the probability of crime of 79 percentage points. This effect persists for a long time span. The model presents an area under the Cumulative Accuracy Profile (CAP) curve, of 0.8 performing similarly or better than state-of-the-art crime prediction models. Public policies seeking to reduce criminal activity and its negative consequences must take into account these mechanisms and the self-exciting nature of crime to effectively make criminal hotspots safer.

Resumen: Efficient allocation of scarce law enforcement resources is a hard problem to tackle. In a previous study (forthcoming Barreras et.al (2019)) it has been shown that a simplified version of the self-exciting point process explained in Mohler et.al (2011), performs better predicting crime in the city of Bogotá - Colombia, than other standard hotspot models such as plain KDE or ellipses models. This paper fully implements the Mohler et.al (2011) model in the city of Bogotá and explains its technological deployment for the city as a tool for the efficient allocation of police resources.

Resumen: El uso del mapeo de puntos calientes (hotspots) es una técnica descriptiva utilizada para identificar el lugar hacia dónde deberían dirigirse los recursos de prevención del crimen. Consiste en la identificación de áreas geográficas donde, en el pasado, ha habido delitos con mayor frecuencia, con el objetivo de prevenirlos en el futuro, de acuerdo con la existencia de determinantes fijos de crimen, así como la existencia de fenómenos temporales que, sin embargo, tienen cierta persistencia. Hay diferentes técnicas para la identificación de los puntos calientes basadas en modelos como el método de estimación de puntos, estimación de densidad por Kernel, el método de elipses espaciales y modelos espacio‐temporales. Este trabajo compara la capacidad descriptiva y predictiva de los principales modelos de predicción de crimen existentes en la literatura académica al caso de crímenes en la ciudad de Bogotá durante el período 2011 – 2012. En términos generales encontramos que el modelo espacio temporal basado en Mohler et. al (2012) es el que mejor desempeño tiene en la predicción del crimen en ciertas zonas de la ciudad