CRIME DATA ANALYSIS
Abstract
In the realm of data science, predictive analytics
has become an invaluable tool for various
sectors, including law enforcement. This paper
explores the development and implementation of
a web application designed to predict the
likelihood of different types of crimes based on
historical data. Utilizing machine learning
models, the application processes inputs such as
the day of the week and specific locations to
provide users with predictions of crime
probabilities. This innovation aims to enhance
crime prevention strategies by offering timely
insights into potential crime occurrences. Crimes
have a negative effect on any society both
socially and economically. Law enforcement
bodies face numerous challenges while trying to
prevent crimes. We propose a Crime Data
Analytic Platform (CDAP) to assist law
enforcement bodies to perform descriptive,
predictive, and prescriptive analysis on crime
data. CDAP has a modular architecture where
each component is built separate from each
other. CDAP also supports plugins enabling
future feature expansions. The platform can
ingest any crime dataset which has the required
attributes to map dataset to attributes required by
the platform. It can then analyze them, train
models, and then visualize data. CDAP also
combines census data with crime data to achieve
more comprehensive crime analysis and their
impact on society. Moreover, with the
combination of census data and crime data,
CDAP provides process reengineering steps to
optimize resource allocations of police forces.
We demonstrate the utility of the platform by
visualizing spatial and temporal relationships in a
set of real-world crime datasets. Predictive
capabilities of the platform are demonstrated by
predicting crime categories, for which a machine
learning approach is used. Identification of
optimized police district boundaries and
allocating patrol beats are used to demonstrate
the prescriptive analytics capabilities of the tool.
Heuristic-based clustering approach was taken
to define police district boundaries in a way
that the identified districts have equitable
population distribution with compact shape. The
resulting districts are then evaluated on
inequality of population and the compactness
using Gini Coefficient and Isoperimetric
Quotient. Another heuristic-based approach was
taken to define new police patrol beats to be
optimized on equitable workload distribution,
compactness, and minimizing response time for
new police patrol beats
Downloads
References
K.Lodha and A.K.Verma, “Spatio-temporal
visualization of urban crimes on a gis grid,” 8th
ACM Intl. symposium on Advances in Geographic
Information Systems, pp. 174–179, 2000.
[2] J. Forgeat. (2015) Data processing
architectures lambda and kappa. [Online].
Available: https://www.ericsson.com/researchblog/
data-knowledge/ data-processingarchitectures-
lambda-and-kappa/
[3] C. Yu, M. W. Ward, M. Morabito, and W.
Ding, “Crime forecasting using data mining
techniques,” 11th IEEE Intl. Conf. on Data Mining
Workshops, pp. 779–786, 2011.
[4] A. T. Murray, I. McGuffog, J. S. Western, , and P.
Mullins, “Exploratory spatial data analysistechniques for examining urban crime implications
for evaluating treatment,” British Journal of
criminology, vol. 41, no. 2, pp. 309–329, 2001.
[5] R. Krishnamurthy and J. S. Kumar, “Survey of
data mining techniques on crime data analysis,”
International Journal of Data Mining Techniques
and Applications, vol. 1, no. 2, pp. 117–120, 2012.
[6] D. E. Brown, “The regional crime analysis
program (recap): a framework for mining data
to catch criminals,” IEEE Intl. Conf. on Systems,
Man, and Cybernetics, vol. 3, pp. 2848–2853,
1998.
[7] H. Chen, D. Zeng, H. Atabakhsh, W. Wyzga, and
J. Schroeder, “Coplink: managing law
enforcement data and knowledge,”
