The San Francisco police department requested a way for visualizing trends in crime data, and The Red Team decided to try to solve the root of their problem. We pulled additional historical records to build a model of decaying peace. Crime may never be removed completely, however we can tell you when peace has been around for a bit too long.
We wanted to model the probability of a specific type of crime appearing in a specific location. Taking the latitude and longitude of SF, we binned the city into multiple geographic buckets. Then we looked for historic trends for crimes in those bucket. We modeled the effect of weekly seasonality on the probability of each crime type, but more interestingly we wanted to know if the streets being too quiet can also have an effect on the probabilities. We ran a logistic regression against weekly seasonality and a linear and quadratic term for the number of days without a crime. Unsurprisingly the higher the number of days without a crime the higher the probability is of a new crime appearing.
We summarized the evolution of the probability of a crime on an animated map.
Built With
- google-cloud
- microstrategy
- postgresql
- python
- r
- redbull
- sql
- tableau
Log in or sign up for Devpost to join the conversation.