Geog-3p95 Cartocrimefinalreport

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

CartoCrime Solutions 2014

Table of Contents
Project Overview

1. Introduction

2. Background

3. Study Area

4 Methodology

10

4.1 Data Preperation

10

4.2 Automation

11

4.3 Data Processing

11

4.4 Variable Data Extraction

12

4.5 Kernel Density

12

4.6 Isolated Crime Variable Cartography (ICVC)

13

5. Results

13

6. Discussion

15

7. Conclusion

18

8. Appendices

19

8.1 - 8.6: Output Maps


8.7 - 8.9: Python Scripts for Automated Processes

9. References

20








2014

CartoCrime Solutions 2014

Project Overview
CartoCrime Solutions is a geospatial company founded in the Niagara Region, Canada, in the
fall of 2014. We are dedicated to providing temporal analyses to crime statistics, aiding by
geospatial technologies. In addition, we produce effective spatial crime information that aids
governments and police agencies globally. Notably, the use of Isolated Crime Variable
Cartography (ICVC) is a tool providing the highest degree of accuracy in relating criminal
activity and social factors connected to similar crime categories. Los Angeles County consulted
CartoCrime Solutions in late 2014 to complete an analysis on criminal activity from 2008
2012. Our work set out to calculate the percentage of change in crime in Los Angeles County
and relay our findings to other social factors linked to criminal behaviour.

1. Introduction

Crime analysis provides potential answers as to where, when and why crimes occur (Phillips and
Lee, 2010). The relationship between the physical layout of an area, proximity to carious
services, and land use mixes are seen as impart factors which are likely to influence criminal
behaviour (Greenburg and Rohe, 1984). Regardless of why crime occurs, criminal behaviour
puts strains on communities and the broader city, as a whole (Murray, McGuffog, Western,
Mullins, 2001). Areas known for criminal activities develop as undesirable locations within
cities. However, crime tends to concentrate around opportunities and other environmental factors
incubating criminal activity (Phillips and Lee, 2010). Analyzing the distribution of crime
connects urban geography to the locations of criminality. It is the ability to combine spatial
information with other data that makes Geographic Information Systems (GIS) valuable for
urban researchers (Murray et. Al., 2001). However, a GIS, only reveals the location of crime (i.e.
hotspot analysis); the utilization of other variables provides explanation as to why crime occurs,

CartoCrime Solutions 2014


opposed to purely a georeferenced point (Murray et. Al, 2001). It is through the combination of
other variables, we see the degree of accuracy increases between a GIS approach to crime
analysis and providing causations for specific types of crimes.
As of 2008, more than half of the worlds human population has resided in urban areas;
and by 2030, urban inhabitants will account for approximately 60 % of the worlds population
(Waibel, 1995). Crime changes perceptions of the quality of life, mental health and physical
security in our daily lives. Hence, having thorough spatial understanding of crime increases in
importance, in a rapidly urbanizing world, ensuring safety for societies. Thus, using geospatial
technologies increases decision-making accuracy in both crime pattern detection and city
planning; serving as a tool to prevent urban crimes from occurring.
Therefore, the following will provide a temporal analysis for the percentage of change in
crime, by zip code, in Los Angeles County from 2008-2012; it will use methodologies such as a
kernel density, comparing changes in crime, to the proximity of four social services (homeless
shelters, substance, abuse services, sheriff and police stations, mental health facilities); create
separate tools, via a python script, to automate the processes of crime location (i.e. clustering)
and identification of single crime types (i.e. violent/nonviolent); and lastly provide a
comprehensive demographic understanding of Los Angeles County to understand other pressures
influencing criminal activity (i.e. Poverty, Unemployment, Mean Household Income).

2. Background Literature
Spatial analysis of crime through use of mapping remains one of the most active fields regarding
technological geospatial applications (Leitner, 2013). Crime within cities is the end result of an
assortment of socioeconomic ills of urbanity, thus crime must be viewed as a collective
assemblage. Through understanding crime as an assortment of variables, it then allows geospatial

CartoCrime Solutions 2014


correlations to be made and analyzed. Unique GIS approaches to crime mapping are grouped
into four fields: fundamental spatial problems, crime analysis, crime modeling, crime mapping,
implementations and applications (Leitner, 2013). Crime mapping allows geographical data to be
transformed into useful information. This information is invaluable to police authorities,
allowing certain trends and phenomena to be studied. Every city has a unique socioeconomic
landscape, method of policing, budget allocation, and crime climate. Therefore, a unique strategy
must be designed for crime prevention within each individual city.
Although digital crime mapping is a recently revealed resource, larger police forces are
adopting computerized crime mapping at rapid rates (Weisburd and Lum, 2005). In 1999 fifty
nine percent of police departments with over one hundred officers within the United States
claimed to use computerized mapping within that year (Weisburd and Lum, 2005). Diffusion of
computerized crime mapping is becoming increasingly popular among police departments;
therefore innovating old approaches to policing. Crime mapping allows police authorities to
develop exceedingly efficient strategies to managing crime, in turn replacing outdated ineffective
methodologies. Also, as demographics and data change from year to year, new-implemented
strategies are open to constructive criticism, leaving ability to be enhanced. Geospatial
technologies permit police establishments to gauge performance of newly implemented tactics.
Through viewing statistics and output data of the digital crime maps, police can identify hotspots
that are actively prone to crime.
Information analysis obtained from crime data is used for a variety of applications,
focusing police distribution more specifically on hot spots, targeting crime prevention efforts
(Levine, 2005). Programs in the past have been developed to further assist police departments in
filtering through data for identifying specific elements to form a criminal spatial synopsis. An

CartoCrime Solutions 2014


example of this type of program is the CrimeStat program. The program is created in C++ and is
multithreading, able to handle large datasets of large metropolitan police departments, and
includes Dynamic Data Exchange to allow other programs to call up CrimeStat and pass the data
name and variable to it (Levine, 2005). There are four versions of this program, with the latest
version of the program updated in 2005. Distribution of crime throughout the city is visually
displayed from the program, allowing clusters to be highlighted and acted upon within reality.
Visually displayed geographical displacement of crime allows police agencies to
understand crime sensitive pockets that are scattered throughout the city. Predictive policing is
an emerging field of police enforcement. If the locations where crimes occur are brought to
attention, police can effectively respond through having a larger police presence distribution in
that locale. As police slash budgets and cannot hire on anymore staff, this nontraditional method
to criminal prevention is becoming an ever-increasingly popular tool (Friend, 2013). Predictive
policing as summarized by Andrew Adams, The goal is to transform policing from a reactive
process, to a proactive process (Greengard, 2012). Police departments in a large number of U.S
cities including Los Angeles, New York, Memphis and Santa Cruz are turning to the concept to
better understand where to allocate resources (Greengard, 2012). Visualizations of crime data are
now possible because of crime mapping, a resource that did not even exist before the late
nineties. Algorithms and mathematical models have been designed to compliment output data
generated by digital crime mapping.

3. Study Area: Los Angeles County & Predictive Policing


The Los Angeles Police Department has worked with UCLA researchers for the past decade to
explore mathematical methods to combat crime (Greengard, 2012). As the LAPD began to
understand that this information was highly influential to police work, it began to make its way

CartoCrime Solutions 2014


into the city on a small scale. Although crime still occurs, strategies developed by the LAPD are
being tested to determine whether or not changes made were indeed effective. Predicting
criminal behavior cannot be a true science, but it allows clarity of city health to be displayed and
acted upon accordingly. Andrea Bertozzi, a professor of mathematics at UCLA says that, crime
will always have a significant stochastic or random component, the non-random component is
what we try and exploit (Greengard, 2012). Pilots for these types of programs are being tested
in metropolitan centers such as Los Angeles containing large datasets, allowing modeling and
analyzing to be based on many incidents of crimes varying in nature.
Map 3.1: Study Area Los Angeles County

Predictive policing is likely to become a cornerstone of law enforcement, allowing for


more effective displacement of people and resources (Greengard, 2012). Data based policing is

CartoCrime Solutions 2014


the way of the future, highlighting trends and phenomena dealing with crime, allowing for a
comprehensive understanding of the city. Los Angeles in being the second largest metropolitan
city within American borders serves as a suitable location to pilot programs pertaining to crime
mapping analysis. In addition, a collection of maps displaying mean household income by city,
unemployment by city, and the poverty rate by city, were compiled to gain a comprehensive
understanding of Los Angeles, County. Through displaying these demographics it allows insight
into the socioeconomic makeup of Los Angeles County and provides community context into the
likelihood of where crimes may be located (Refer to Map 3.2: Mean Household Income, Los
Angeles County/ Map 3.3: Unemployment Rate Los Angeles County/ Map 3.4: Poverty Rate
Los Angeles County); these maps are examples of statistical demographic cartography.
Map 3.2: Mean Household Income Los Angeles County

CartoCrime Solutions 2014


Map 3.3: Unemployment Rate Los Angeles County


















Map 3.4: Poverty Rate Los Angeles County

CartoCrime Solutions 2014

4. Methodology

4.1: Data Preparation
The first step concerning data preparation is to create a File Geodatabase to store all of the work
for the project, because File Geodatabases "are ideal for storing and managing geospatial data
(Childs, 2009). Following creation of the File Geodatabase, multiple comma separated value
files

of

all

crimes

from

2008

and

2012

were

downloaded

from

this

link:

http://shq.lasdnews.net/CrimeStats/CAASS/desc.html. The zip code boundaries were then


downloaded from this link provided by the Los Angeles County government data portal:
http://egis3.lacounty.gov/dataportal/2011/01/24/zip-codes-with-parcel-specific-boundaries/.
Since this data was acquired as a shapefile, it was simply exported to the Geodatabase as a
feature class called 'crime_master'. The city boundaries were also downloaded from the Counties
data portal, the link for this is: http://egis3.lacounty.gov/dataportal/2014/06/18/city-boundaries/.
This data was also acquired as a shapefile, therefore it was simply exported to the Geodatabase
as a feature class called City Boundaries. Excel documents were then downloaded
encompassing information concerning unemployment rates mean household incomes, and
poverty rates by city from:
http://www.dof.ca.gov/research/demographic/state_census_data_center/american_community_s
urvey/. After all of the raw data was acquired, a process of data modification was implemented
to refine our output data. Extraneous data was then removed from excel spreadsheet such as:
records representing cities in other counties, unneeded fields, and unincorporated areas of Los
Angeles (city boundaries Feature Class did not contain any records for these areas). Abbreviation
of the City Name field was altered to remove the word 'city' from the end of each city name;
allowing for data to be joined to the City Boundaries Feature Class. Cells were then formatted to

10

CartoCrime Solutions 2014


be 'number' or 'general' data type as opposed to custom, which is not recognized by ArcMap.
Each excel document was then saved as a .CSV file allowing junction to the City Boundaries.
4.2: Automation
A python script was developed that automates the task of converting a CSV file to a feature
class. This script makes a temporary XY Event Layer based on the user-input CSV file and then
writes it to a feature class. An additional python script was created to count the number of
crimes in each zip code area. This script first constructs a new field within the zip code areas
feature class. An update cursor then selects and counts the number of crimes in each postal code
area and writes the counted values to the new field. An option is also available to the user
displaying an SQL expression that would have the update cursor count crimes only from a
certain category. Construction of the third python script allows the user to calculate the percent
change in crime between two years for each zip code area. This script simply creates a new field
to contain the percent changes in crime, and then calculates it based on two user input field
names from the same feature class that represent two different years of crime data. These python
scripts were developed to work with one another to turn raw, unprocessed data into meaningful
results. The user is only required to have a boundaries feature class, which is usually widely
available as parts of open GIS datasets, and a CSV file containing co-ordinates of crimes, which
could be obtained from even the cheapest of GPS units.
4.3: Data Processing
The first python script was executed on each of the crimes CSV files to create a feature class for
crimes in 2008 and 2012. The second python script was also executed twice to count the number
of crimes in each zip code area for each year. The third python script was then executed on the
field created by the second python script in order to calculate the percent change in crime from

11

CartoCrime Solutions 2014


2008 to 2012 within each zip code area. This attribute was symbolized and mapped accordingly
to delineate zip code areas with high increases in crimes. In order to demonstrate the capabilities
of the second script, it was executed on the 2008 and 2012 crime feature classes to count only
drug related crimes. The third python script was executed to calculate the percent change in drug
related crimes from 2008 to 2012. This attribute was mapped underneath a semi-transparent
kernel density of substance abuse centers to pinpoint areas with high increases in drug related
crimes and few substance abuse centers. The city boundaries feature class was joined with each
of the CSV files for poverty rate, unemployment rate, and mean household income. The joins
were based on a common city name field. After the joins were executed, each variable was
symbolized and mapped to delineate the different social standing of each city.
4.4: Variable Data Extraction (Social Service Layers)
The available geodatabase provided a single feature class, containing thousands of records
outlining the current social services located within Los Angeles County. Vast amounts of the
data were not relevant for crime analysis. Therefore, drug abuse centers, mental health facilities,
sheriff and police stations, and lastly, homeless shelters were queried by their attributes from the
dataset.
4.5: Kernel Density
Once the four separate social services were exported into their own feature classes, a kernel
density method was chosen in ArcGIS. This spatial analyst tool was used to create a raster layer,
using colour values to represent the spatial density magnitude, per square mile for each social
service. After experimenting with the tool multiple times, in an effort to understand how large of
a raster grid cell size would suffice to visualize the appropriate density, we concluded on a cell
size of 500 by 500 pixels.

12

CartoCrime Solutions 2014


Next, we created a density raster layer for each social service (4 in total); these rasters
were overlayed on top of the percent change of crime feature class, in each zip code area, from
2008 to 2012. Preparing these data sets allows spatial analysis to be conducted. As a result, we
develop an understanding of how these variables are related to crime and ultimately, how they
can benefit municipal and police planning applications moving forward.
4.6: Isolated Crime Variable Cartography (ICVC)
In addition, our python script [determining the percentage change for crime in each zip code
boundary] was further developed to isolate individual types of crime. Granting the capability to
compare our social service density layers, to a particular (i.e. related) crime. Moreover, by
adding an SQL query into our script, it grants the end-user capability to select particular crimes
of interest. Thus, allowing precise investigation accuracy in identifying spatial trends in crime.
For example, drug related crime percent changes, per zip code, were compared to the spatial
densities of drug abuse centers. As a result, isolated crime variable cartography (ICVC) provides
the end-user with greater understanding of the spatial dynamics of particular crime variables,
opposed to simply examining the total crime trends per zip code.
Furthermore, this project combines temporal analysis, kernel density, isolated crime
variable cartography (ICVC) and statistical demographic cartography, to obtain optimal accuracy
in crime pattern recognition and connections between related crimes.

5. Results
5.1: Percent of Change in Crime by Zip Code (Appendix Map 8.1)
This map allows a visual representation of crime percent change to be displayed. Areas that are
darker in color signify a rising occurrence in crime, while lighter colors such as white represent

13

CartoCrime Solutions 2014


areas that have not changed or have decreased in crime. Construction of this output map allows a
base map for kernel density overlay to be possible.
5.2: Kernel Density Police and Sheriff Station Density (Appendix Map 8.2)
Overlay of raster data regarding police and sheriff station density is placed on top of the crime
percent change map to identify areas that lack police presence. Dark areas that lack police
presence can then be investigated by authorities to determine whether or not resources should be
redistributed to a specific location, to prevent further criminal activity. Geographical
displacement of police officers may be altered depending on the conclusions drawn from crime
experts.
5.3: Mental Health Facility Density (Appendix Map 8.3)
Mental health facility locations are displayed as an overlay to percent change in crime to
determine the effects of lack of services on crime rates. Mental health facilities provide stability
to the mentally unstable, therefore having a correlation on total crime rates.
5.4: Substance Abuse Program Density (Appendix Map 8.4)
Substance abuse program density was overlaid against crime percent change to determine where
lack of services was located. Analysis and algorithms can then be constructed to determine areas
that should incorporate more substance abuse centers.
5.5 Homeless Shelter Density (Appendix Map 8.5)
Spatial displacement of homeless shelters is visually presented against percent change in crime
to determine whether lack of homeless shelters affects crime rates.
5.6: Change in Drug Related Crime by zip code vs. Substance Abuse Centers - (Appendix
Map 8.6)

14

CartoCrime Solutions 2014


Drug related crime was specifically extracted from the data to compare a correlated variable.
Areas that experience low density in substance abuse centers and high in crime percent change
require more services to compensate.

6. Discussion
6.1: Percent of Change in Crime by Zip Code (Appendix Map 8.1)
Although there are areas within this map that signify a drop in crime, there are many areas that
undergo drastic spikes in crime. Some of these values exceed 20,000 percent rise in crime in
certain areas, allowing the most vulnerable locations to be pinpointed. Dark areas are then able to
undergo in depth analysis when compared against other socioeconomic variables.
6.2: Police and Sheriff Station Density - (Appendix Map 8.2)
From viewing this map it is apparent that clustering of police and sheriff stations are
concentrated within the central inner city. From the center of the cluster to the southwest portion
of the map the color yellow is displayed. This signifies that density of police and Sherriff stations
in this area are lower then the urban core. There are many areas that have elevated percentage
crime change in crime that do not have a police station within an immediate jurisdiction. The
areas that have zero or purple density with elevated crime rates are crime sensitive and require
more police enforcement.
6.3: Mental Health Facility Density (Appendix Map 8.3)
This map signifies that much dense clustering of mental health facilities occur within the core of
the inner city. The green values represent this. Sprawling out from this central location there is a
lot of purple values, meaning that there is low density of homeless shelters. Also, within these
low-density areas there are considerable amounts of elevated crime percent change. Areas

15

CartoCrime Solutions 2014


experiencing low density in mental health facilities and high percent change in crime make for
crime sensitive hot spots.
6.4: Substance Abuse Program Density - (Appendix Map 8.4)
Within the inner city moving south there is three clusters of areas that include high density of
substance abuse programs. These areas are experiencing less crime then the surrounding areas
that do not contain substance abuse programs, therefore presence of programs has a positive
correlation with reducing crime. Many areas to the North do not contain any resources regarding
substance abuse, and signify a high percent change in crime. Resources must be redistributed to
enhance quality of life in these areas.
6.5: Homeless Shelter Density (Appendix Map 8.5)
This map presents one cluster of homeless shelters within the inner city. Sprawling out in all
directions, there is an immediate decline in density of homeless shelter presence. As the density
declines the crime percent rises, allowing potential social infrastructure locations to be built to
lower crime rates within the city.
6.6: Change in Drug Related Crime by zip code vs. Substance Abuse Centers - (Appendix
Map 8.6)
Much of the map is raided with red, meaning that there is a low density of substance abuse
centers located throughout the city. Within the inner city moving south there is three clusters of
areas that include high density of substance abuse programs. In relation to crime percent change
the cluster of shelters represented in green are in the correct areas. With this being true, there are
also many areas that lie to the north and to the west of this location that require more substance
abuse centers. Areas that contain red or no color for density that have a high percent change are
the areas that require socioeconomic restructuring.

16

CartoCrime Solutions 2014

6.7: Limitations
Crimes that are on a zip code boundary get counted as occurring in both coded areas. This is
roughly ~150-500 crimes that get duplicated. This occurs as a result of the operator using the
'Select by Location' part of the second script.
'WITHIN' operators give same result.

It was determined that the 'INTERSECT' and

The 'COMPLETELY WITHIN' operator was not a

suitable alternative because it excludes the crimes on a boundary. Some postal code areas have
1000%+/- change in crime, which might be attributed to missing data. It is quite possible that
these numbers are skewed as a result of jurisdiction constraints. The link where the crime data
was downloaded from states that although the Sheriff provides the data, the Sheriff does not have
jurisdiction over the entire county of Los Angeles. Consequently, if jurisdiction ever changed or
if the sheriff became more/less involved in making arrests, the data will be skewed. Crimes
without XY coordinates are not included in the results. Although some of these crimes have
postal codes associated with them, this is simply a data quality issue - the important thing is that
for good results, good input data must be used. The crimes without XY data but with a postal
code don't seem to exist within the postal code areas with drastic % changes in crime, meaning
that finding a way to include the crimes with only a zip code may not help our analysis. Again,
this all relates back to worthy input data is synonymous with credible output data. Initially, we
wanted to use crime data from 2005, 2009, and 2013. However, the data that we found for
unemployment mean household income, and the poverty rate dated from 2008-2012. Our tools
allowed us to simply download the corresponding crime data, and process it quickly.

17

CartoCrime Solutions 2014

7. Conclusions
Further research would improve the creditability of this approach to crime analysis. In particular,
increasing temporal dimension provides a broader understanding of the increases or decreases in
criminal activity in geographic locations. Also, providing understanding into the need or lack of
need, in developing more social infrastructure to combat crime. Merging additional socioeconomic data increases the accuracy and understanding of cities, and more specifically,
neighbourhoods within city boundaries. Lastly, in future development, CartoCrime Solutions
looks to incorporate a web-GIS application to make this data available to the public, offering the
democratization of criminal data to the masses.
Furthermore, the following analysis has provided evidence into the validity of
incorporating geospatial technologies in relation to crime. Our study illustrates calculating
percent changes in crime, enhances municipalities understanding of criminal activity, and thus,
provides insight into the appropriate allocation of funding and resources in relation to crime
spending/prevention. For example, the City of Los Angeles in 2012 spent $84,783,737
concerning crime control (Budget Summary, City of Los Angeles). Integrating our crime
analysis techniques increases the accuracy for stakeholders in municipal planning, via the
production of budget summaries, location of police presence and lastly, the placement of social
service infrastructure, in combating the likelihood of crime occurrence.
CartoCrime Solutions use of isolated crime variable cartography (ICVC) is a method
granting greater understanding between particular crimes and related social phenomenon. As the
literature states, GIS simply reveals the location of crime (i.e. hotspot analysis); utilizing other
variables provides explanation as to why crimes occur, opposed to a georeferenced point.
(Murray, McGuffog, Western, Mullins, 2001). Thus in order the fill literature gap, ICVC,

18

CartoCrime Solutions 2014


enhances end-users ability to determine suitable causations for particular crimes, simultaneously,
gaining geographic understanding of crime distribution. In addition, police forces across the
globe, only require the use of X and Y coordinates, for our automation process to calculate not
only the percent of change in crime, but the physical location. Subsequently, the use of ICVC or
a kernel density of external variables increases police forces ability to protect our most
vulnerable neighbourhoods. On a final thought, the use of CartoCrime Solutions tools in
conjunction with the analyses of social factors influencing crime furthers municipalities spatial
understanding of crime in cities.

8. Appendices
8.1: Percent of Change in Crime by Zip Code

8.2: Police and Sheriff Station Density

8.3: Mental Health Facility Density

8.4: Substance Abuse Programs Density

8.5: Homeless Shelter Density

8.6: Change in Drug Related Crime by zip code vs. Substance Abuse Centres

8.7: Percent Change Python Script

8.8: Crime Counter Python Script

8.9: CSV to Feature Class Python Script

19

CartoCrime Solutions 2014

9. References
City of Los Angeles. (2012). 2012 Fiscal Budget Summary. Los Angeles, California.

Friend, Z. (2013) Predictive Policing: Using Technology to Reduce Crime. FBI Law
Enforcement Bulletin.
Greenburg, S. and Rohe, W. (1984). Neighbourhood Design and Crime. Journal of the American
Planning Assoication: 50: 48-61.
Greengard, S. (2012) Policing the Future. West Linn: Society
Phillips, P., and Lee, I. (2011). Crime analysis through spatial areal aggregated density patterns.
Geoinformatica: 15: 49-74.
Levine, N. (2004). Crime Mapping and the CrimeStat Program. Houston: Ohio State
University.

Murray, A., McGuffog, I., Western J., and Mullins, P. (2001). Exploratotry spatial data analysis
techniques for examining urban crime. Centre for Crime and Justice Studies: 41: 309329.
Singleton, A and Brunsdon, C. (2013). Escaping the pushpin paradigm in geographic
information science: (re)presenting national crime data. Liverpool: University of
Liverpool.

Waibel, M. (1995). Drawn and quartered Hanois ancient centre is feeling the pressure from
outside and in, So what can architects and officials do to stop the rot? Vietnam Econ.
Times: 24-25.

Weisburd, D and Lum, C. (2005) The Diffusion of Computerized Crime Mapping in Policing:
Linking Research and Practice. Manassas : George Mason University

20

You might also like