Model validation:

Homicide prediction is a challenging task due to the spatiotemporal sparsity of these criminal events. In this paper, we report the results of using several approaches to mitigate this sparsity condition in machine learning models specifically designed to model homicide events. Since spatial resolution is a direct determinant of sparsity, we focus on the performance of these models at different resolutions of interest to law enforcement. We use a simple count model as a benchmark and propose some improvements aimed at improving prediction performance. We then compare the results with more complex models motivated by multiple learning methods and graphical signal processing. We find that simple benchmark models are as good as state-of-the-art models for low resolution, but, as the resolution increases, the performance of the machine learning models outperforms the benchmark. These results provide a rationale for the use of state-of-the-art machine learning models for homicide prediction at the high resolution of interest for police resource deployment.