Integrated Safety Incident forecasting and Analysis

The proposed effort is a part of a global city action cluster team with the Nashville Fire Department and Metro Nashville Police Department. Both in prior research, as well as practice, incident forecasting and response are typically siloed by category and department, reducing effectiveness of prediction and precluding efficient coordination of resources. Advances in data collection and storage infrastructure offer an unprecedented opportunity to integrate both data and emergency resources from distinct urban agencies into a dynamical learning system anticipating and rapidly responding to heterogeneous incidents, such as distinct categories of crime, as well as car accidents. The goal of this research is to develop methods for continuous-time spatio-temporal incident forecasting using both crime and fire response incident data, and to develop decision support tools that leverage such forecasts to optimize both resource allocation and response under uncertainty. In particular, the resulting optimization framework would determine which units to dispatch (police, fire, or both) in order to minimize expected response time, and what equipment they should bring, taking into account the time, location, and nature of incidents as they arrive, as well as those predicted to occur in the future.
The proposed research faces a number of technical challenges requiring significant advances in fundamental research. The key algorithmic challenge on the forecasting side is continuous-time forecasting of spatio-temporal time series of heterogeneous incidents. In tackling the forecasting task, we propose to develop methods to cluster incidents in space, as well as by category, and use the resulting clustering to develop distinct continuous-time models to forecast incident arrival distribution based on survival analysis. The optimization framework, in turn, requires a scalable solution for integrated spatio-temporal allocation of heterogeneous emergency responders, making use of developed integrated forecasting methods. The proposed optimization methods will transform the incident response problem into a transportation problem with heterogeneous resources, which can be formalized as a network-flow linear program, augmented to account for heterogeneity in the resources and incidents that these resources can address. This subproblem will then be embedded in an iterative Bender's decomposition to account for the potential of certain categories of resources, such as police vehicles, to actually impact the distribution of incidents (in this case, crimes).
The broader impact of this research is its application in the context of a global city teams challenge action cluster. During this work, the decision support and analytical tools we develop will be disseminated to the community. More- over, the research has the potential to impact actual operational planning at the Metro Nashville Police Department and Nashville Fire Department, by integrating the data sources and optimally coordinating response. Broader im- pacts also include involvement in educational activities, including lectures for High School students at the School for Science and Math at Vanderbilt, active engagement of undergraduates in research, and undergraduate and graduate teaching. Finally, the project will provide support for two Ph.D. students.