Content uploaded by Paulo Joao
Author content
All content in this area was uploaded by Paulo Joao on Feb 03, 2016
Content may be subject to copyright.
The Geography of Criminality
– Information and GIS models
for Decision Support
PAULO JOÃO
DEIO – Statistics and Operational Research Department- FCUL
JORGE FERREIRA
e-GEO, Research Centre for Geography and Regional Planning, FCSH – UNL
JOSÉ MARTINS
e-GEO, Research Centre for Geography and Regional Planning, FCSH – UNL
Abstract: Over the last few years a new worldwide socio-economical
order lead to an increasing number on crime rates and raised the need
to find new ways to handle information about criminality. To better
understand its causes, local, regional and national security authorities
turned to new decision support tools such as Geographic Information
Systems (GIS) and other information technologies to help them in finding
better solutions. To understand the true magnitude of all the variables
involved it was necessary to spatially capture and correlate them to
better quantify and qualify the context within the phenomena. The city
of Lisbon with is new proposed administrative division, reducing from
53 to 24 “freguesias” (minimum administrative division and similar to
parish’s) implies an enormous degree of uncertainty in the observation
and location of criminal data. As the crime is not treated with an exact
point, but at the level of parish, it implies that larger parishes are treated
by the average crime regardless of place of occurrence. This research
combines statistical methods (cluster analysis) and spatial models created
with GIS based on police crime reports. It also details a framework for
short-term tactical deployment of police resources in which the objective
is the identification of areas where the crime levels are high (enough) to
enable accurate predictive models as well as to produce rigorous thematic
Politeia
Ano IX – 2012, pp. 145-164
146 POLITEIA – Revista do Instituto Superior de Ciências Policiais
maps. In recent years police services have engaged on proactive and
Intelligence-Led Policing (ILP) methods. This advance was coincident
with the recognition of law-enforcement solutions at local level. This
paper is also an approach to ILP as a strategic methodology to provide
tools for Decision Support System (DSS) by police departments.
Keywords: Crime analysis, GIS, Geostatistics, Intelligence-led Policing,
Information Dissemination, Data Mining.
1. Introduction
Over the last few years a new worldwide socio-economical order
lead to an increasing number on crime rates and raised the need to find
new ways to handle information about criminality. To better understand
its causes, local, regional and national security authorities turned to new
decision support tools such as Geographic Information Systems (GIS)
and other information technologies to help them in finding better solu-
tions. To understand the true magnitude of all the variables involved it
was necessary to spatially capture and correlate them to better quantify
and qualify the context within the phenomena.
The city of Lisbon with is new proposed administrative division;
reducing from 53 to 24 parishs (minimum administrative division) im-
plies an enormous degree of uncertainty in the observation and location
of criminal data. As the crime is not treated with an exact point, but
at the level of parish, it implies that larger parishes are treated by the
average crime regardless of place of occurrence.
In recent years police services have engaged on proactive and
Intelligence-Led Policing (ILP) methods. This advance was coincident
with the recognition of law-enforcement solutions at local level.
This research combines statistical methods (cluster analysis) and
spatial models with GIS based on police crime reports. It also details
a framework for short-term tactical deployment of police resources in
which the objective is the identification of areas where the crime levels
are high (enough) to enable accurate predictive models as well as to
produce rigorous thematic maps.
It is also an approach to “Intelligence-led policing” as a strategic
methodology to provide tools for decision support by police departments.
Recent advances on geostatistics, information, data mining and GIS,
the study and modeling of crime data to identify patterns has emerged as
147
The Geography of Criminality – Information and GIS Models
a new research field. Within the scope of this paper, some key questions
will be addressed:
-
cincts accurately enough for use in deployment, scheduling for
evaluating police effectiveness?
new patterns?
to foresee and predict the occurrence of incidents, beyond the
mere descriptive models?
2. Geographical Information Systems and Crime Analyses
Since the 1960’s Geographical Information Systems (GIS) have
been applied to a vast number of studies and criminality is not an ex-
ception. From its first applications in Canada, GIS has become a major
instrument for an effective territorial planning.
From in-car navigation, retail and commerce location, costumer
geo-marketing studies to risk management, construction, weather fore-
casting, military planning and other application fields.
However, it was only in the beginning of the 1980’s with the
reductions in the price of technologies (operating systems, processors,
storage capacity, memory and hardware), that GIS saw a significant
development on new research areas, such as crime analysis, distribution
of police precincts and planning crime reduction.
Much of the work in crime mapping and analyses was carried out
in the United States by the Mapping and Analysis for Public Safety
(MAPS), formerly known as the National Institute of Justice’s Crime
Mapping Research Center (CMRC).
This work served as a launching platform to the development of
crime mapping in other countries like South Africa, Australia and United
Kingdom.
The computerization of police records has come with a realization
that this material can be used for crime and intelligence analyses (Ra-
tcliffe, 2004). This work permits the recognition of patterns sometimes
hidden and often not perceived by the police and authorities.
Geography is also necessary for advances on spatial understanding
and consequently “…has contributed to many disciplines where unders-
148 POLITEIA – Revista do Instituto Superior de Ciências Policiais
tanding space and place is important, such as with crime” (Ratcliffe,
2004).
With the new Lisbon administrative division, crime analyses will
be more difficult due to area aggregation. For example on certain types
of crime, aggregation will probably mask important pattern details. The
comparison between the new and the old administrative divisions in Lis-
bon is an important and the use of predictive models can help authorities
and provide the necessary results to an effective decision.
To proceed or not with the new administrative division is the major
question that has to be answered shortly. And the most correct answer
will depend upon a rigorous and pertinent research based on information
and knowledge.
The comparison between the new and the old administrative divi-
sions in Lisbon is an important exercise to perceive pattern differentiation
and has obvious implications on territorial planning and public domain.
The use of geostatistics tools such as predictive models and kriging
methods allows, beyond the mere description of phenomena, to foresee
and predict the occurrence of incidents.
3. Intelligence-Led Policing (ILP)
Police forces always struggling with a lack of human resources try
to reduce crime on their jurisdiction by analyzing and correlate infor-
mation from primary sources on criminal environment.
Criminal historical data are analyzed to develop new strategies
and actions to reduce crime. It is a very demanding task in terms of
technological means and human resources but it is the best strategy to
reduce crime because it has an approach, both preventive and repressive.
Figure1: ILP and crime reduction process (Ratcliffe, 2005)
Criminal environment
Decision-maker
Intelligence
149
The Geography of Criminality – Information and GIS Models
An ILP model at its first stage enables the interpretation of the
criminal environment. This is usually performed by an intelligence sec-
tion, and relies on the range of information’s sources both within and
external to the police service.
The obtained information should, in an intelligence-led envi-
ronment, be passed to people who can actually impact in a positive
manner on the criminal environment (decision maker). This requires an
intelligence structure to be able to identify and influence the decision-
-makers. It should be noted that this requires both an ability to identify
the decision-makers, as well as to influence their thinking regarding the
types of strategies to achieve better criminal numbers.
ILP strives for greater efficiency in policing, but it has also been ac-
companied by other efficiency methods, some of which conflict with ILP.
There is a performance culture in many police services which stri-
ves to measure everything possible, and it is a concern that the benefits
of ILP will be lost in a flood of operational statistics.
One of the major issues is response time. Many police services have
now records of the response time to routine calls, and they build response
improvement’ protocols that maximize performance. Unfortunately the
research evidence is fairly conclusive: improving response times to calls
for service does not reduce crime.
The intelligence led policing model, is more difficult to implement
but is also the most versatile in performance and results.
The following table compares the various models of policing in
terms of implementation, resources and results:
150 POLITEIA – Revista do Instituto Superior de Ciências Policiais
Table 1: Strategic aspects of police activity
Implementation
dificulty
Human resources
Technology
resources
Costs
estimate
Geography
application
Expected
results
Type of
information
Type of
actuation
Neiborhoud
and commu-
nity policing
low high low high Low low retrospective preventive
Problem
oriented
policing
medium high medium high medium medium retrospective repressive
Intelligence-
led policing
high low high medium High high prospective preventive
and
repressive
In recent years there has been a move within police services towards
a proactive and ILP. This has coincided with recognition of the value of
local policing solutions and the importance of the intelligence function
at the local area. This paper is an approach to ILP as a strategic metho-
dology to provide a tool for the decisions makers in police departments.
Knowing where the police become a crucial intervenient element
provides a better understanding of the intervention program and anti-
cipated outcomes and effects. Without such a baseline, evaluations are
limited as to the purposes and sometimes anticipate consequences to
police efforts. Intelligence-led policing is also about being a smart police.
That means, increasing the effectiveness and efficiency of police
interventions through the use of the right information. Such efforts
are data driven (quantitative and qualitative) and analytically rigorous,
assuring than appropriate care has been taken to collect, organize and
analyze information, prior to action.
This ameliorates the deficiencies of rigid boundaries and moves
towards a more dynamic methodology. With the new Lisbon paradigm,
from 53 to 24 parishes and assuming these administrative limitations,
crime analysis will be more difficult due to the aggregation.
151
The Geography of Criminality – Information and GIS Models
4. Crime Mapping Today
With the emergence of geographic information technology and its
successful use as a tool for crime analysis and forecasting, the study
and modeling of historical or current crime data to identify spatial crime
patterns has emerged as a new research area. However in Europe, this
reality is still recent. Three key questions can be pointed out:
patterns in the short-run small areas such as precincts accurately?
as counterfactuals for evaluating police effectiveness?; and
forecast models that can predict turning points and other new
patterns, such as the start of a new crime?
For some crime types, aggregation may mask important seasonal
variations in smaller ecological areas. For example, university areas may
have seasonal patterns in property crimes influenced by the comings
and goings of the students and teachers. Also, shopping areas may have
peaks by holiday times.
Law enforcement officers have been mapping crime virtually since
the time that police agencies were established through the use of push
pins and a paper map at the beginning of the 20 Century (Harries, 1999).
By the late 70’s and early 80’s, there was a resurgence of interest
in understanding and analyzing the spatial dimension of crime through
the characteristics of incident, its location and geographical analysis on
incident (Boba, 2009). With the development of GIS in the 90’s, law
enforcement officers begun to consider the principles of geography and
spatial information using new technologies in order to realize and deve-
lop mechanisms to crime prevention and the public safety.
The cartographic representation has some advantages such as: help
visual analysis and statistics of crime, aggregate information in spatial
matrix, produce thematic maps that help to communicate the results of
the analysis (Eck, 2005 & Boba, 2009).
Harris (1990:3), states that “the purpose of maps is to communi-
cate information”. The cartographic representation is a tool to commu-
nicate information across time and space.
The diffusion of GIS into crime analysis has been a slow process
primarily because of cost (both hardware and software) and complexity.
152 POLITEIA – Revista do Instituto Superior de Ciências Policiais
When discussing the current use of GIS for crime analysis, activities can
be divided into current and cutting edge applications (currently used in
law enforcement agencies to support crime analysis activities).
The uses of GIS to support crime analysis activities in police day-to-
-day operations are a powerful tool for strategic planning (Johnson, 2000).
4.1. Crime Hot-Spots
This is the most common method used in criminal representation.
It assumes that the locations of concentration criminal past will persist
into the future, however the actual results of this method depends on
the time period under review, usually this robust method only produces
good results when applied to short time series (Spellman, 1995).
To Adman’s-Fuller’s (2001), an interesting feature in the detection
of hot-spots is its persistence and coincidence over time as shown by
Anselin (2000), the hot-spots reflect high levels of crime initially mode-
rate, but over time, usually, this crime will change to more violent types
of crimes (e.g. acts of vandalism to crimes of theft).
Therefore, should be contained and controlled in time to prevent
more serious incidents to people and property in the geographic area
covered by the hot-spot.
According to Eck (2005), this method assumes that they must map
the locations and not criminal occurrences thus understand why certain
settings are more easily criminal occurrences while others appear to inhi-
bit these same events. Ainsworth (2001:88) refers that “A crime hot-spot
is usually understood as a location or small area with boundaries clearly
identified where there is a concentration of criminal incidents, which
exceed the normal for this area, the term can also be used to describe
locations that showed an increase in crime a given period of time”.
4.2. Descriptive Statistics for Crime Analyses
One of the most important tasks associated with analyzing criminality
is know your data, this data are the baseline for an intelligence analyze, ho-
wever, very little of this data are collected with intelligence analysis in mind.
Understanding the different types of data and their definitions is
important because some types on analyses have been designed for par-
ticular types of data and may be inappropriate for another type.
153
The Geography of Criminality – Information and GIS Models
After preprocessing and coding all the information, data was
analyzed and the dataset consisted in 35549 police records, distributed
by 8 variables, according with the fallowing coding:
Table 2: Data coding process
Variable Min Max
Day 1 31
Month 1 12
Hour 0 23
Minute 0 59
Type of crime 1 6
Subtype of crime 1 34
Classification 1 112
Parish 1 53
According to the Persons’ coefficient correlation matrix, there was
no significant correlation between the variables witch let us to identify
the variables as independent observations.
Table 3: Person correlation matrix
Day Month Hour Minute Type Subtype Classification Parish
Day 1
Month -0,011 1
Hour 0,008 0,003 1
Minute 0,007 0,000 -0,019 1
Type 0,050 0,016 0,013 -0,020 1
Subtype 0,004 0,002 0,003 -0,004 0,020 1
Classification -0,014 0,009 0,010 -0,009 -0,008 0,000 1
Parish -0,002 0,005 0,007 -0,013 0,012 0,020 0,032 1
154 POLITEIA – Revista do Instituto Superior de Ciências Policiais
In terms of information, the data represented a small part of the
criminal reality, since there were only considered the crimes reported
and recorded by the Public Safety Police (Polícia de Segurança Pública
– PSP).
The primary dataset consist of one year of criminal offences for all
individual events for 2009 in Lisbon County obtained with authorization
from the PSP.
The dataset include all personal and patrimonial offences. The most
often studied crime is robbery occurred in public transportation. All the
records has a parish of reference, which permits the classification of the
offences and their mapping in a GIS.
Unfortunately the accuracy is not the ideal because we don´t have
the points of the occurrences which prevents us from using the geore-
ference point but only the “parish area”.
The histograms below show that the majority of crimes are robbery
especially during the evening.
Figure 2: Data histograms
4.3. Methodology
Regarding the GIS methodology adopted, it was necessary to
prepare the work with a set of layers and create a theme to match the
expected distribution of parishes. Thematic maps showing the number
of crimes or other metrics were aggregated using color palettes (darker
for more crimes and brighter for less).
T
inLisbo
n
Thedat
a
occurre
d
classific
a
U
occurre
n
Thehist
o
Figure2
4.3.Me
t
R
layersa
n
thenu
m
andbrig
T
thecri
m
isfairb
reality.
T
useofr
a
Thistyp
e
delimita
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
Density
T
heprimary
n
Countyob
t
a
setinclude
d
inpublic
a
tionofthe
o
U
nfortunat
e
n
ceswhich
p
o
gramsbelo
:Datahisto
g
t
hodology
R
egardingt
h
n
dcreatea
m
berofcrim
e
hterforless
T
hismetho
d
m
ehasmore
yMunicipal
T
hemainre
a
a
tiosinther
e
e
ofmapsc
r
tions(orjur
i
02
datasetco
n
t
ainedwith
a
allpersona
l
transportati
o
ffencesan
d
e
lytheacc
u
p
reventsusf
wshowtha
t
g
rams
h
eGISmeth
o
themeto
m
e
sorother
m
).
d
iswidelyu
s
impact.Ho
w
ity(e.g.Lis
b
a
sonisbeca
e
presentati
o
r
eatesajuri
s
i
sdictions),
w
46
Crimetype
n
sistofone
y
a
uthorizatio
n
l
andpatri
m
on.Allthe
d
theirmapp
u
racyisnot
romusingt
h
t
themajorit
o
dologyad
o
m
atchtheex
m
etricswer
e
s
edbecaus
e
w
ever,altho
u
b
on)andit
usethepop
o
nofcolors
w
s
dictionalco
n
w
hentheya
r
8 10
y
earofcrimi
n
n
fromtheP
m
onialoffen
c
recordsha
s
inginaGIS.
theideal
h
egeorefere
yofcrimes
a
o
pted,itwas
pecteddistr
e
aggregated
e
itispercei
v
u
ghthisisal
canleadt
o
ulationden
s
w
ouldbeth
e
n
flictbetwe
e
r
edifferent(
12
n
aloffences
SP.
c
es.Themo
s
s
aparish
o
becausew
e
ncepointb
u
a
rerobbery
e
necessaryt
ibutionofp
usingcolor
v
ed,forexa
m
soapopula
r
o
interpreta
t
s
ityisdiffer
e
e
wisestcho
i
e
nthegeog
r
asisthecas
e
forallindiv
i
s
toftenstu
d
o
freference
e
don´thav
e
u
tonlythe“
p
e
speciallyd
u
opreparet
h
arishes.Th
e
palettes(d
a
m
plebythe
r
method,it
t
ionsthatd
o
e
ntineachp
i
ce.
r
aphicalare
a
e
ofLisbon).
i
dualevents
d
iedcrimei
s
,whichpe
r
e
thepoin
t
p
arisharea”
.
u
ringtheev
e
h
eworkwit
h
e
maticmaps
a
rkerformo
r
county,tho
s
isunlikelyt
h
o
notcorre
s
lace.Inthis
a
sandadmi
n
for2009
s
robbery
r
mitsthe
t
softhe
.
e
ning.
h
asetof
showing
r
ecrimes
s
ewhere
h
atcrime
s
pondto
case,the
n
istrative
131
155
The Geography of Criminality – Information and GIS Models
This method is widely used because it is perceived, for example
by the county, those where the crime has more impact. However, al-
though this is also a popular method, it is unlikely that crime is fair by
Municipality (e.g. Lisbon) and it can lead to interpretations that do not
correspond to reality.
The main reason is because the population density is different in
each place. In this case, the use of ratios in the representation of colors
would be the wisest choice.
This type of maps creates a jurisdictional conflict between the geo-
graphical areas and administrative delimitations (or jurisdictions), when
they are different (as is the case of Lisbon).
Figure 3: Parish’s administrative division (actual and planning)
On crime analysis, cartograms were initially created to show ab-
solute values in crime (recorded in 2009) by the Public Security Police
(PSP) in Lisbon, according to the current structure of 53 parishes and
the new structure of 24 parishes.
This representation was the result of the geo-referenced crimes re-
corded in 2009 by 53 existing parishes. To represent the crimes recorded
132
Figure3:Parish’sadministrativedivision(actualandplanning)
Oncrimeanalysis,cartogramswereinitiallycreatedtoshowabsolutevaluesincrime
(recordedin2009)bythePublicSecurityPolice(PSP)inLisbon,accordingtothecurrentstructureof
53parishesandthenewstructureof24parishes.
Thisrepresentationwastheresultofthegeo‐referencedcrimesrecordedin2009by53
existingparishes.Torepresentthecrimesrecordedbythenewadministrativestructurewas
necessarytotheaggregatecrimedatabynewparishes.TheparishofSantaMariadosOlivais,under
thenewstructurewillbedividedintotwoparishes,givingrisetoanewone,theOriente.
Inordertoestimatethevalueofapproachingcrimeinthesetwoparishes,itwasnecessarytodivide
thetotalbythenewareaofcrime,multiplyingthecurrentarea.Theresultwasthensubtractedfrom
theoldparishwithaviewtoobtainingtheaveragevalueestimatedforthisnewdelineationoftwo
parishes.
Itshouldbenotedthatthesedatacouldnotbeexactbecauseasmallerareamayhavemore
orlessrecordedcrimes.However,forresearchpurposes,asinthiscase,themethodcanbeused,
providedthatisproperlyreferredto.
156 POLITEIA – Revista do Instituto Superior de Ciências Policiais
by the new administrative structure was necessary to the aggregate crime
data by new parishes. The parish of Santa Maria dos Olivais, under the
new structure will be divided into two parishes, giving rise to a new
one, the Oriente.
In order to estimate the value of approaching crime in these two
parishes, it was necessary to divide the total by the new area of crime,
multiplying the current area. The result was then subtracted from the
old parish with a view to obtaining the average value estimated for this
new delineation of two parishes.
It should be noted that these data could not be exact because a
smaller area may have more or less recorded crimes. However, for re-
search purposes, as in this case, the method can be used, provided that
is properly referred to.
Figure 4: Total Crimes in Lisbon
Analyzing crime data by the actual divisions and by the new ones,
it seems obvious an increase in numbers (in part due to the aggregation
of values recorded in several parishes that lead to new administrative
area). Importantly, the low number of crimes in the small villages of
133
Figure4:TotalCrimesinLisbon
Analyzingcrimedatabytheactualdivisionsandbythenewones,itseemsobviousan
increaseinnumbers(inpartduetotheaggregationofvaluesrecordedinseveralparishesthatlead
tonewadministrativearea).Importantly,thelownumberofcrimesinthesmallvillagesofthe
currentdivision,risingfromfirsttothelastclass(whenaggregatedthenewadministrativedivision).
ThedatashowsaverylargecriminaldensityonOriente,buttheempiricalknowledgeofthisareacan
tellusthatmostcriminalincidentsoccurinits(new)aggregatedneighborhood.
Next,weproceededtocalculatethecriminaldensity,fortheparish,dividingthetotalcrimesper
squareKminarea.
157
The Geography of Criminality – Information and GIS Models
the current division, rising from first to the last class (when aggregated
the new administrative division).
The data shows a very large criminal density on Oriente, but the
empirical knowledge of this area can tell us that most criminal incidents
occur in its (new) aggregated neighborhood.
Next, we proceeded to calculate the criminal density, for the parish,
dividing the total crimes per square Km in area.
Figure 5: Ratio of Total Crimes per Square Km in Lisbon
Regarding criminal density, the new administrative division gives
the perception of a reduction in crime numbers. However, this fact is
in part related to the significant increase in the area of some parishes,
thus reducing the density of crimes per square Km.
In order to understand how it would be the spatial representation
of crime according to the city’s population, created a criminal incidence
rate, based on dividing the number of crimes by the resident population,
the results were expressed in ‰.
The data source for the population was raised in the Fourteenth
general population census conducted by the National Statistical Institute
134
Figure5:RatioofTotalCrimesperSquareKminLisbon
Regardingcriminaldensity,thenewadministrativedivisiongivestheperceptionofa
reductionincrimenumbers.However,thisfactisinpartrelatedtothesignificantincreaseinthe
areaofsomeparishes,thusreducingthedensityofcrimespersquareKm.
Inordertounderstandhowitwouldbethespatialrepresentationofcrimeaccordingtothecity's
population,createdacriminalincidencerate,basedondividingthenumberofcrimesbytheresident
population,theresultswereexpressedin‰.
ThedatasourceforthepopulationwasraisedintheFourteenthgeneralpopulationcensus
conductedbytheNationalStatisticalInstitute(INE)in2001,wouldbemorecredibletousevaluesfor
thepopulationin2009,butitwasnotpossibleduetothefactthatthecensusbeconductedonly
from10to10years,andpopulationestimatesrefertotheadministrativeunitabovethecity.
158 POLITEIA – Revista do Instituto Superior de Ciências Policiais
(INE) in 2001, would be more credible to use values for the population
in 2009, but it was not possible due to the fact that the census be con-
ducted only from 10 to 10 years, and population estimates refer to the
administrative unit above the city.
Figure 6: Ratio of Criminal Incidence
per 1000 Inhabitants in Lisbon
In relation to the criminal incidence rate, the new administrative
division gives the perception of a reduction in crime recorded even
more significant, comparing with the result obtained by calculating the
density criminal.
The reduction is due in part to the significant increase in the area
of some parishes and their residents, thus reducing the incidence rate of
crime in certain districts of the new administrative division.
It is known that the southern part of the city, corresponding to the
downtown district, correspond to an area that has the highest number
of recorded crimes, mainly because it is the center of major economic,
financial and commercial activities. But also because it’s the main tou-
rism “spot” in the city, which triggers criminal activity.
135
Figure6:RatioofCriminalIncidenceper1000InhabitantsinLisbon
Inrelationtothecriminalincidencerate,thenewadministrativedivisiongivesthe
perceptionofareductionincrimerecordedevenmoresignificant,comparingwiththeresult
obtainedbycalculatingthedensitycriminal.
Thereductionisdueinparttothesignificantincreaseintheareaofsomeparishesandtheir
residents,thusreducingtheincidencerateofcrimeincertaindistrictsofthenewadministrative
division.
Itisknownthatthesouthernpartofthecity,correspondingtothedowntowndistrict,
correspondtoanareathathasthehighestnumberofrecordedcrimes,mainlybecauseitisthe
centerofmajoreconomic,financialandcommercialactivities.Butalsobecauseit’sthemaintourism
“spot”inthecity,whichtriggerscriminalactivity.
Anenormouslackofresidentpopulationmustalsobeconsideredwhenanalyzingthedata.
Withtheaggregationofexistingparisheswithverydifferentrates,thereisadecreaseinthehigh
incidencerateinthecriminality
Inordertoachievedifferentperspectiveswhenanalyzingthecrimeresultsacrossthecity,itwas
decidedtomakethesumoftheresidentpopulationwiththemobilityofpopulation(populationthat
entersandcirculatesinthecityandtourists),inthesensethatonlybyanalyzingthecriminal
statisticsaccordingtotheresidentpopulationofLisbon,maybetooverstatetheintensityof
occurrence.Assumingthatalargenumberofresidentpopulationandthismobilitymayleadtoan
increasedlikelihoodofacriminalcase(Machado,2007b:138).
159
The Geography of Criminality – Information and GIS Models
An enormous lack of resident population must also be considered
when analyzing the data. With the aggregation of existing parishes with
very different rates, there is a decrease in the high incidence rate in the
criminality
In order to achieve different perspectives when analyzing the crime
results across the city, it was decided to make the sum of the resident
population with the mobility of population (population that enters and
circulates in the city and tourists), in the sense that only by analyzing
the criminal statistics according to the resident population of Lisbon,
may be to overstate the intensity of occurrence. Assuming that a large
number of resident population and this mobility may lead to an increased
likelihood of a criminal case (Machado, 2007b:138).
Figure 7: Ratio of Criminal Incidence per 1000 Inhabitants
and Presents in Lisbon
Looking at the crime rate with the local population and this, it is
possible to observe the emergence of new parishes with high rate of
crime, however, the maximum crime rate is much lower when compared
to the representation only by the resident population.
136
Figure7:RatioofCriminalIncidenceper1000InhabitantsandPresentsinLisbon
Lookingatthecrimeratewiththelocalpopulationandthis,itispossibletoobservethe
emergenceofnewparisheswithhighrateofcrime,however,themaximumcrimerateismuchlower
whencomparedtotherepresentationonlybytheresidentpopulation.
Finally,itwascreatedaspatialindexbasedoncrimeincidenceusingspatialstatisticaltools.Thistype
ofoperationshouldbeperformedatamicroscaleofanalysis,statisticallyequaltosubsection
(Martins,2010),constitutingthehighestlevelofdisaggregationcorrespondingtotheblockinurban
terms(Geirinhas,2001).
ThisindexwascalculatedusingaGeographicWeightedRegression(GWR)usingspatial
contextofdynamicGaussianKernel.Thisoperationusesthesquarerootorlogofthevariablesthat
areselectedforanalysisbysmoothingtheabsolutevaluesandapproximatingthecurveofnormal
distributioninparabolicform,reducingdisparitiesinthedistribution.
Theresultisexpressedinstandarddeviationunits,whichisrepeatedlyusedasanindexof
risk(Harries,1999).GWRprovidesalocalmodelofthevariableorprocessyouaretryingto
understand/predictbyfittingaregressionequationtoeveryfeatureinthedataset.GWRconstructsa
separateequationforeveryfeatureinthedatasetincorporatingthedependentandexplanatory
variablesoffeaturesfallingwithinthebandwidthofeachtargetfeature.
TheshapeandextentofthebandwidthisdependentonuserinputfortheKerneltype,
Bandwidthmethod,DistanceandNumberofneighborsparameters
160 POLITEIA – Revista do Instituto Superior de Ciências Policiais
Finally, it was created a spatial index based on crime incidence
using spatial statistical tools. This type of operation should be performed
at a micro scale of analysis, statistically equal to subsection (Martins,
2010), constituting the highest level of disaggregation corresponding to
the block in urban terms (Geirinhas, 2001).
This index was calculated using a Geographic Weighted Regression
(GWR) using spatial context of dynamic Gaussian Kernel. This operation
uses the square root or log of the variables that are selected for analysis
by smoothing the absolute values and approximating the curve of normal
distribution in parabolic form, reducing disparities in the distribution.
The result is expressed in standard deviation units, which is repe-
atedly used as an index of risk (Harries, 1999). GWR provides a local
model of the variable or process you are trying to understand/predict by
fitting a regression equation to every feature in the dataset. GWR cons-
tructs a separate equation for every feature in the dataset incorporating
the dependent and explanatory variables of features falling within the
bandwidth of each target feature.
The shape and extent of the bandwidth is dependent on user input
for the Kernel type, Bandwidth method, Distance and Number of nei-
ghbors parameters
The dependent variable for the GWR index, consisted on the cri-
minal occurrences and as an independent variable, it was used the sum
of the resident population with the mobility of population.
It should be noted that this type of analysis can be a part of nu-
merous independent variables of socio-economic development in order
to identify a more realistic index.
However, with an area of study at the parish, it would be a difficult
assignment of certain socio-economic values in a larger scaled area.
It also worth mentioning that this type of fee (if used alone without
performing a normalization of the variables) may generate incorrect
analysis results (Harries, 1999) due to the fact that there are high levels
of occurrence in areas with low population and this happening also with
a cartographic representation of crime rates.
161
The Geography of Criminality – Information and GIS Models
Figure 8: Spatial Index of Criminal Incidence per 1000 Inhabitants
(actual and planning)
Looking at the previous representation and using the GWR ope-
ration, it seems evident the reduction of higher values of distribution
according to population, mainly in certain areas of the new administra-
tive division.
5. Conclusions and Future Works
The final results of this research can be vital to decide an effective
police strategy. The decision to proceed with this new administrative
scenario is so difficult as to decide about new forms to deal with this
new territorial sketch. To decide what bases to consider and what tech-
nologies to adopt are also difficult choices.
In recent years, researchers and technicians have made huge pro-
gresses in harnessing the analytic capabilities of GIS to track crime
patterns over time and then, use this information to create predictive mo-
137
ThedependentvariablefortheGWRindex,consistedonthecriminaloccurrencesandasan
independentvariable,itwasusedthesumoftheresidentpopulationwiththemobilityofpopulation.
Itshouldbenotedthatthistypeofanalysiscanbeapartofnumerousindependentvariablesof
socio‐economicdevelopmentinordertoidentifyamorerealisticindex.
However,withanareaofstudyattheparish,itwouldbeadifficultassignmentofcertainsocio‐
economicvaluesinalargerscaledarea.
Italsoworthmentioningthatthistypeoffee(ifusedalonewithoutperforminga
normalizationofthevariables)maygenerateincorrectanalysisresults(Harries,1999)duetothefact
thattherearehighlevelsofoccurrenceinareaswithlowpopulationandthishappeningalsowitha
cartographicrepresentationofcrimerates.
Figure8:SpatialIndexofCriminalIncidenceper1000Inhabitants(actualandplanning)
LookingatthepreviousrepresentationandusingtheGWRoperation,itseemsevidentthe
reductionofhighervaluesofdistributionaccordingtopopulation,mainlyincertainareasofthenew
administrativedivision.
5.ConclusionsandFutureWorks
Thefinalresultsofthisresearchcanbevitaltodecideaneffectivepolicestrategy.The
decisiontoproceedwiththisnewadministrativescenarioissodifficultastodecideaboutnewforms
162 POLITEIA – Revista do Instituto Superior de Ciências Policiais
dels. These advances turned GIS in a valuable tool to assist and support
decision-making strategies for the police forces and security services.
Now that many law enforcement agencies have adopted crime
mapping and have begun to produce the types of tools mentioned, they
want more.
The demands for more sophisticated spatial analytic techniques
lead to the research on predictive models to help the prevention of the
“next crime”.
The crime must be examined in the context of threats and trans-
forms the societies themselves, taking into account assessment of the
demographic, economic, social and environmental that could affect the
criminal act itself. Just seeing the real threats and developing solutions
and using Geography, GIS and multivariate statistics analysis, we can
analyze the current map of criminality.
The advanced spatial analysis came to identify crime patterns and
vulnerable areas (in terms of insecurity). This type of analysis facilitates
knowledge in order to make strategic decisions in to combat the criminal
phenomenon.
However, only at micro scale, the criminal data can be considered
as useful to a strategic planning against crime.
Complementing spatial analysis of crime with the empirical know-
ledge of the historical and cultural components of a given territory, one
may identify the reason for the occurrence of a particular type of crime
in a given geographical area and therefore planning and resources in
order to prevent and reducing it.
Nowadays, the level of spatial analysis and modeling techniques
already permit the development of tactical planning strategies to fight
crime. Data gathered in a form of information and properly worked
can end up in an effective decision-support tool. And that can only be
achieved by a combination of technology and human knowledge.
References
ainsWortH, P. (2001), Offender Profiling and Crime Analysis, UK, Devon,
William Publishing.
anselin, L. et al. (2000), Spatial Analyses of Crime, Criminal Justice: Vol. 4,
USA, Washington.
Bação, F. (2006) Introdução ao Data Mining. Lisboa, ISEGI-UNL.
Bailey, D. (1994) Police for the future, Oxford University Press, New York.
163
The Geography of Criminality – Information and GIS Models
BloCk, R. (1995) Space, place and crime, New York, USA: Criminal Justice
Press.
BoBa, R. (2009), Crime Analysis with Crime Mapping, SAGE Publications,
Inc, California, 2nd Ed.
CHainey, S.; Thomson, L. & Uhlig, S. (2008), The Utility of Hotspot Mapping
for Predicting Spatial
Patterns of Crime, in Security Journal, Vol. 21, No. 1-2. (February 2008), pp.
4-28.
eCk , J. et al. (2005), Mapping Crime: Understanding Hot Spots, USA, National
Institute of Justice.
Ferreira, J. (2004), A Geografia da Sociedade da Informação em Portugal,
PhD Thesis, FSCH-UNL p.555.
FilBert, K. (2008), Targeting Crime in Hot Spots and Hot Places, Geography
& Public Safety, vol 1, Issue 1, February 2008, Washington, DC, COPS,
NIJ, pp.4-7.
GroFF , E.; et al. (2001) Forecasting The Future of Predictive Crime Mapping.
(n. l.): Crime Prevention Studies, Volume 13:29-57.
Harries, K. (1999), Mapping Crime: Principle and Practice, Washington, DC,
NIJ.
João, P. (2009) Modelo Preditivo da Criminalidade – Georeferenciação ao Con-
celho de Lisboa, MSc. Dissertation, ISEGI-UNL, Lisbon. New University
of Lisbon, MSc Dissertation available at http://run.unl.pt/handle/10362/3424
JoHnson, C. (2000), Crime Mapping and Analysis Using GIS. Geomatics
2000: Conference on Geomatics in Electronic Governance, Pune, Geomatics
Group, Center for Development of Advanced Computing, Pune University
Campus, p.5.
JonatHan, J. et al. (2003) Predicting the geo-temporal variations of crime and
disorder. Glamorgan, UK: Elsevier.
Julião, R. (2001), Tecnologias de Informação Geográfica e Ciência Regional:
Contributos Metodológicos para a Definição de Modelos de Apoio à Decisão
em Desenvolvimento Regional, FCSH-UNL, Lisbon.
kate , B.; et al. (2004) Prospective Hot-Spotting – The Future of Crime
Mapping? in The British Journal of Criminology, vol. 44 num. 5: Oxford
University.
kellerMan, A. (2002), The Internet on Earth: A Geography of Information,
John Wiley e Sons, England.
Longley, P. et al. (2005), Geographic Information Systems and Science, Wiley
Blackwell, 2nd Ed.
MaCHado, P. et al. (2007b), Metrópoles Seguras: Bases para uma Intervenção
Multisectorial nas Áreas Metropolitanas de Lisboa e Porto, 2 Volume, Lis-
boa, LNEC – MAI.
164 POLITEIA – Revista do Instituto Superior de Ciências Policiais
Martins, J. & Ferreira, J. (2009), A Geografia do Crime, paper presented
at the VII Congresso de Geografia Portuguesa, 27 de Novembro de 2009,
Universidade de Coimbra.
Martins, J. (2010), Os Sistemas de Informação Geográfica na Gestão da
Criminalidade em Portugal, Msc. dissertation FCSH-UNL, Lisboa.
Mielke, P. (2008b), Creating Base Maps and Layer Files for Cartographic
Consistency, Geography & Public Safety, vol. 1, Issue 2, July 2008, Wash-
ington, DC, COPS, NIJ, pp.9-11.
ratCliFFe, J. & McCullagh, M. (2001) Chasing Ghosts – Police Perception of
High Crime Areas. British Journal of Criminology, 41:330-341.
ratCliFFe, J.H. (2004).Strategic Thinking in Criminal Intelligence. Sydney:
Federation Press.
ratCliFFe, J.H. (2005). The effectiveness of police intelligence management:
A New Zeland case study. Police Practice and Research, 6 (4).
Wilson, R. & Paulsen, D. (2008), Foreclosures and Crime: A Geographical
Perspective, Geography & Public Safety, vol. 1, Issue 3, October 2008,
Washington, DC, COPS, NIJ, pp.1-2.
Wilson, R. & Smith, K. (2008), What is Applied Geography for the Study
of Crime and Public Safety?, Geography & Public Safety, vol. 1, Issue 1,
February 2008, Washington, DC, COPS, NIJ, pp.1-3.
Wortley, R. & Mazerolle, L. (2008), Environmental Criminology and Crime
Analysis: Situating the Theory, Analytic Approach and Application, William
Publishing, Devon, UK, pp.1-18.
