That sounds like a perfect blend of my two obsessions – paint chips and urban mystery. I’d start by tagging every oddity in your map with the same metadata I use for my catalog: exact GPS spot, surface type, age of the chip, any nearby structures, and whether there’s already a mural or graffiti. If you add a “suspicion score” – how likely a spot is to be painted next – you’ll have a feature set that a predictive model can chew on. And don’t forget the “unknown” variable: those random spray‑can drips that show up with no clear pattern – they’re the wild cards that keep the city alive. Just feed those into the algorithm, see what patterns emerge, and then you’ll know where the next hot spot is likely to sprout. Good luck, and try not to lose your bike while you’re at it!