bensoltoff
commented
6 years ago - Predictive analytics for crime - its been done before. Look for previous research tackling this problem to see what existing strategies are and build from there
- How detailed will your predictions be? Zip code level? Census tract level? Block level?
- Contrasting case studies - a good entry point
- How much data cleaning will need to occur? Do the two cities use the same reporting system and database structure?
- Have you taken classes with Luc Anselin? He does tons of spatial data work, might be a good reference for this paper
- With two separate cities, be sure to make similar comparisons. That is, you are measuring or predicting the same underlying construct/phenomena in both cities
- Are any of these algorithms specifically tuned for temporal or spatial data structures?