SafeWalk is an app that finds the safest route between locations at any time of day. It uses NYC open crime data from 2005-2016, the Dijkstra algorithm to find shortest paths, and a cost model that assigns crime costs based on type and time. The app provides crime avoidance level and personal fear of crime ratings to help users choose routes. It was created by a physicist studying solitons and is validated using third party crime data.
2. Data Sources and Pipeline
NYC Open Data
Major Crimes
Dataset
2005-2016
(~1,000,000)
3. Algorithms
● Dijkstra Algorithm: between Start and End, find
the shortest path minimizing sum of weigths
● K-nearest neighbours classification to associate
each crime with a road
d1
d4
d5
d2
d9
d3 d6
d8
d10
d12
d7
d11
Start
End
4. Cost Model for Road Lengths
Personal fear of crime Crime Costs depending
on hour and type of
crime
● De Sota Road, NY at 0300
5. App Landing Page
● Personal Fear of Crime: Rational, Almost Rational,
Borderline Rational-Irrational, Irrational, Extreme
6. Brooklyn Safe Route at 0200
● Crime Avoidance Level : Vigilant, Personal Fear of Crime:
Borderline Rational-Irrational
7. Cost Model Validation (3rd
Party)
● Courtesy of Trulia
https://www.trulia.com/local/new-york-ny/tiles:1|
points:0_crime
8. Personal Fear of Crime: Rational
● Crime Avoidance Level: Vigilant
9. Personal Fear of Crime: Extreme
● Crime Avoidance Level: Vigilant
10. About Me
● Physicist studying solitons in
photonic crystal waveguides
● Motorcycle enthusiast
12. Personal Fear of Crime
● Personal fear of crime affects each route slightly
differently but there exists some saturation point
13. Safety Rating
● approximates probability of experiencing no
crime on route
● Probablity of crime on road i.
● Assume crimes on each road are
independent