Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

What's the time function on your graph? Are you only showing the last X weeks or all time? Things like police patrols could affect the frequency of crime over time, and that may change your predictions and confidence intervals.

EDIT: Depending on your data freshness, you may want to do something like push notifications if a crime happened in your area in the past 4 hours. That way if someone walks through a "good area" but there's recent crime they should be more vigilant.



We managed to gather data for about a month when building this app. Some of them were as recent as a few hours and some a few days.

We implemented a weighted average algorithm to calculate the risk levels and how confident we were of the data. For the risk levels I took into account 1) How recent the data was 2) What time of the day it occurred relative to the time of query (for eg: the crime occurred at 4pm on another day and its 5pm now, it will be given a higher weightage compared to something that happened at say, 11am.) 3) The severity of the crime itself (murders >> theft) 4) Number of occurrences of crimes in the area

The assignment of weights wasn't thoroughly scientific though. It needs some tweaking.

And thanks for the suggestion on push notifications :)




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: