Series

Predicting Crime in Aotearoa

8 parts · all series

  1. Intro

    Predicting Crime in Aotearoa

    NZ Police publish over a million crime records openly. What happens when you point deep learning at them?

    5 min
  2. Part 01

    Wrangling a Million Crime Records

    NZ Police's crime dataset is publicly available, but it's UTF-16 encoded, full of trailing periods, and 32% of records don't know what time the crime happened. Here's how we cleaned it up.

    4 min
  3. Part 02

    Giving Crime a Place on the Map

    Crime records come with names and codes but no coordinates. Here's how we joined 1.15 million records to Stats NZ boundary files and gave every crime a place in physical space.

    4 min
  4. Part 03

    Crime as Video

    Turn a million geo-tagged crime records into a 4D tensor by overlaying a 500m grid on Auckland. Crime prediction becomes video prediction.

    4 min
  5. Part 04

    What the Data Actually Shows

    Before throwing deep learning at Auckland crime data, you need to actually look at it. Seasonal patterns, spatial hotspots, and the sparsity problem.

    4 min
  6. Part 05

    Can You Beat Last Month?

    Before deep learning gets a chance, we need to know how well stupid-simple models perform. Turns out, they put up a real fight.

    3 min
  7. Part 06

    Teaching a Neural Network to Watch Crime Like Video

    ConvLSTM was built for weather radar. Turns out predicting crime on a grid is basically the same problem. Here's how it works and what it learned.

    6 min
  8. Part 07

    Three Ways to Look at Time

    ST-ResNet decomposes crime patterns into three temporal scales and models each one separately. Clever architecture, but does it actually help with only four years of NZ data?

    6 min