There are a number of other potential avenues for further research in this domain. Most obviously, there is a need to test the performance of the algorithms in alternative settings: in other cities and countries, and for other crime types. In addition to this, however, there is broad scope for the development of algorithmic variations. In this work, we chose a relatively simple predictive algorithm—a kernel-based approach—in order to place the emphasis on the network context (and to avoid confounding the comparisons made). The general framework, however, could be applied to any predictive method, including others which have performed well in grid-based implementations. The adaptation of these algorithms, and the development of novel approaches tailored to the network context in particular, represents an exciting direction for further work.