2. Problem Statement
● Is the shortest route always the “safest” route?
● How does the route depend on time of day,
type of crime?
3. Data Sources and Pipeline
NYC Open Data
Major Crimes
Dataset
2005-2016
(~1,000,000)
4. Algorithms
● K-nearest neighours classification to associate
each crime with a road e.g. Robbery at edge 8.
● Dijkstra Algorithm: between Start and End, find
the shortest path minimizing the sum of weigths
w1
w4
w5
w2
w9
w3 w6
w8
w10
w12
w7
w11
Start
End
5. KISS Model for Road Lengths
Personal fear of crime Crime Costs depending
on hour and type of
crime
● De Sota Road, NY at 0300
Actual road length
Effects of personal bias
Crime Avoidance Level: Vigilant
Crime Avoidance Level: Neutral
6. App Landing Page
● Personal Fear of Crime: Rational, Almost rational, Borderline irrational, Irrational,
Paranoid
7. Brooklyn Routing at 0100
● Crime Avoidance Level : Neutral, Personal Fear of Crime: Rational
11. Insights
● NYC is pretty safe at any time of time of day
(Nice! but not for me).
● One’s perceived ‘fear’ of crime can lead to
significant route modificationss
● There exists a trade-off between route length
and safety
13. Demos!
● Brooklyn neutral Show time of day
● Port Harris Neutral vs Paranoid Show personal
fear of crime parameter
●
14. About Me
● What I do besides solving (numerically) partial
differential equations.
●
15. KDE
● Cool figure of crime density goes here
●
● : number of crimes (type i) on road/total
number of crimes (type i) in precinct
● :
● : total number of crimes on road/total number
of crimes in precinct