Hi, this is Wayne again with a topic “How predictive policing software works”.
People have been drawing maps since the early 1800s to try and understand where and when crime happens in the hopes of better predicting where and when it will happen again. Today’S crime maps are made by computers and fed by detailed crime data police departments routinely scan their precincts for hot spots places experiencing a disproportionate amount of criminal activity. With the help of these maps, many departments simply put cops on the dots in order to prevent future crimes. This hotspot analysis has become more sophisticated, leading to predictive policing. Most predictive policing begins with a grid dividing a city into squares that can be as small as half a block beginning with data of previous crimes. Computer programs create a forecast for each square, estimating the risk of a crime occurring there during the precincts next four-hour shift. But beyond that, predictive policing systems differ in their approaches. Pred poll was one of the first predictive policing tools on the market and its forecasts are based solely on crime data. What kind of crime happened where and when it builds on techniques for forecasting earthquakes with the aftershocks of the crime reverberating across the city’s grid, then there’s risk terrain modeling, which draws on the same basic crime data as well as features of the city like proximity to A public transportation hub, a collection of bars in the current weather or even the phases of the Moon punch, lab, uses the same advanced factors as risk terrain modeling and adds a learning algorithm a tree based technique for assigning risk to grid cells.
You can think of this. Like a game of 20 questions. Have there been complaints about disorderly conduct in nearby recently? Is there concentration of bars or liquor stores nearby? Is there a subway stop nearby answering each question: sends you down a different branch of the tree? The end of the process being a risk value ultimately, hunch lab, creates hundreds of these trees and combines them into a forest to create a robust forecasting tool. Will we ever really be able to predict crimes before they happen? Maybe not, but even the idea prompts important questions about how we manage police resources.
.