ANALYZING AN OFFENDER’S JOURNEY TO CRIME
USING A CRIMINAL MOVEMENT MODEL
Presenters
Andre Norton
Karen Lancaster-Ellis
What is the main goal?
The main goal of this study is to model the
movement of offenders journey to crime in
order to investigate the relationship between
their road network travel route and the actual
locations
of
their
geographic space.
crimes
in
the
same
Brief Historical Overview
Trinidad– 1864 sq. m
Tobago – 116 sq. m
Est. Population-1.3m
Ethnicity – Cosmopolitan
Religions – Predominantly
Muslim, Hindu & Christian
Politics- Democracy
Police Stations & Posts - 77
Strength – Approx. 6,500
Economy- Petro-Chemical sector
Key Words
KEY WORD
MEANING
Node
Home, Recreation, Entertainment, Work (locations).
Activity Space
An area familiar to an individual through his/her through
everyday activities such as where they live, work, commutes or
goes shopping;
Crime Attractor
A location that attracts offenders
opportunity for crime;
Crime Generator
A location which attracts a large amount of people without and
which presents opportunities to commit crimes (e.g. a shopping
mall);
because of its known
Awareness Space Route individuals traverse to and from typical locations of
activity in their daily lives;
Self-Containment
Index
The percentage of crimes in an area that is committed by
offenders who live in the area;
Background
Crime Triangle
Adapted from Clarke & Eck (2003)
Overlapping “Activity Space”
Previous Research

Consensus across Academic community – Journey to
crime research
 Offender don’t tend to travel to far to commit crime


Major weakness discovered



Distance decay pattern
Sole focus on distance to crime
Distance in isolation (ambiguous results)
Shortcoming Addressed

Our research addresses all 3 dimensions
 Starting point
 Direction travelled
Introduction
Research Title

Environmental Criminology (Brantingham & Brantingham, 1990)

Criminology & Computer Science
 Crime
Pattern Theory
 Dijkstra’s
Algorithm (Edgar Dijkstra, 1956)
Crime Pattern Theory

Criminals – preferred areas to commit crimes

Criminal events likely to occur

Activity space of offenders overlap with activity space of victims

Activity space of target is simply its location
(Brantingham & Brantingham, 1990, Felson & Clarke,1998)

Theory- The social intervention level

Nature & Immediate situations in which crime occurs
Dijkstra’s Algorithm
G (V,E)
Some uses of Dijkstra’s Algorithm ?

Urban traffic planning;

Optimal pipelining of VLSI chip;

Telemarketer operator scheduling;

Routing of telecommunications messages;

Network routing protocols (OSPF, BGP, RIP);

Optimal truck routing through given traffic.
Computer Science- Graph Theory
Environmental Criminology
Offender & Victim Activity
Space example
Vertex / Node
A – B – C – D = 28
A – F – E - D = 29
A – B – D = 22
A – C – F – E - D = 26
A – C – D = 20
Study Area

Four Attractor Locations – Southern Division
 Gulf
City Mall- (largest shopping mall in City)
 Teddy’s
 Space
 Pizza
Shopping Center
La Nouba (Social Night Club)
Hut Food Complex
*** See depiction on next slide
Study Area
Entertainment
Activity Space
Buffer Zone
Home
Recreation
Recreation
Awareness Space
Methodology
Data Sets- 2006-2010 crime data (Serious Crimes)
 Anonymised
 Offence
crime reference number (GO)
type (robbery, vehicle theft)
 Anonymised
geographic co-ordinates (crime location &
offender residence)
 Offender’s
 Date,
name, date of birth, home address
time and location (each offence committed)
 Estimated
monetary value of goods stolen
Methodology
Data Sets- 2006-2010 crime data (Serious Crimes)



Defaults to text file format (.txt)
Converted to .csv & imported into Microsoft Excel
Manual pre-processing to improve data quality
Data Integrity- Significant number of offenders recorded multiple
addresses (This caused some issues)
 Home node – Single address
Distinct offender addresses – Table A
Methodology
Table A
Crime Type
Crime
Trips
Offenders
Offender
Addresses
Robbery
90
300
654
Larceny Motor Vehicle
60
130
346
Weaknesses
• Present in virtually all published studies that use data of a
similar source or nature
Analyses – Directly comparable with previous work (extends)
Dijkstra’s Model Requirements
INPUTS

Spatial data – road network connecting attractors

Distance Measures (length of road networks)

Temporal data- Estimated travel time (impactors)

Encoded into adjacency Matrices (plot network)

Crime address- Where committed

Offender address- Home address

Attractor location- towards which offenders travel
Assumptions
Scenario
Assumption
Shorter distance from crime Offender travelling in direction
location to attractor than from of attractor;
home;
Crime location in direction of Attractor location coded as
specific attractor;
Crime
attractors;
location
offender destination;
between Closest attractor assigned as
potential crime location;
Dijkatra’s Pseudocode
Function Dijkstra’s (Graph, source)
for each vertex v in Graph
//Initialization
dist[v] := infinity
//Initial distance from source to vertex v is set to infinity
previous[v] := undefined
//Previous node in optimal path from source
dist[source] := 0
//Distance from source to source
Q := the set of all node in the Graph // All nodes in the graph are unoptimized thus are in Q
While Q is not empty:
//Main loop
u := node in Q with smallest dist[]
remove u from Q
for each neighbor v of u:
//Where v has not been removed from Q
alt := dist[u] + dist_between(u,v)
if alt < dist[v]
//Relax (u,v)
dist[v] := alt
previous[v] := u
return previous
Experimentation


Octave program – High level programming language

Inputs imported into Octave (Matlab compatible)

Model run based on inputs (1hrs)
Dijkstra’s Algorithm – Time efficiency

worst-case running time O (n2) – Input size

Model took 2.5hrs with Dijkstra’s shortest path calculations
Results
Results
Analysis (Crimes)
Attractor Locations Crimes
80%
Activity Space Crimes
20%
Model generated paths
55% simulated the offender crime location
Self-Containment Index
90%
Euclidean Measure
30% offenders used
Dijkstra’s Measure
70% Offenders used shortest Path
Predictive Analytics
25% (approximation)
Impact & Implications
• Situational Crime Prevention
• Policy Formulation
• Evidence Based Approach (Sherman, 1998)
• Predictive Policing – Data Mining & Big Data solutions
Conclusions & Future Research Priorities

Test bed for geographic profiling of volume crime;

Robbery offences (include victim’s JTC in model);

Increase the number of offences being analyzed;

Eliminate assumption of home node start point;

Increase the geographical size of the analyzed study area.
Acknowledgement
THE END
Andre.Norton@ttps.gov.tt (1-868-469-8523 / 489-4556)
Karen.Lancaster-Ellis@ttps.gov.tt (1-868-2950 / 489-5382)