-
01
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?
6min 1,490 w -
11
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.
5min 1,376 w -
21
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.
3min 797 w -
31
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.
4min 990 w -
41
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.
4min 944 w -
51
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.
4min 1,071 w -
61
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.
4min 868 w -
71
Predicting Crime in Aotearoa
NZ Police publish over a million crime records openly. What happens when you point deep learning at them?
5min 1,267 w