Page 105 - Proceedings of the 2017 ITU Kaleidoscope

FOSTERING SMART CITY DEVELOPMENT IN DEVELOPING NATIONS: A CRIME
SERIES DATA ANALYTICS APPROACH

Omowunmi E. Isafiade, Antoine B. Bagula

Department of Computer Science,
University of the Western Cape, South Africa.

ABSTRACT smart city goal focused at improving the “quality and effi-
ciency” of the services rendered by government entities and
Crime remains a challenge in many parts of the world. This decision makers [3].
is compounded in low-resource settings where police are
short-staffed and there are not enough technological solu- Security agencies (particularly in developing nations) need
tions in place to assist security agencies with knowledge- to adopt more reliable and promising crime mining solutions
driven decision support. While most smart city initiatives to realise better tactical and strategic ways of dealing with
have placed emphasis on the use of modern technology such crime. One of such ways is through the timeous identifica-
as armed weapons for fighting crime, this may not be suffi- tion of a crime series pattern (CSP) as depicted in Figure 1,
cient to achieve a sustainable safe and smart city in resource which is a set of crimes considered to have been committed
constrained environments, such as in Africa. In particular, by the same offender [4]. However, our findings reveal that
crime series which is a set of crimes considered to have been police do not currently have an automated means of identify-
committed by the same offender is currently less explored in ing CSP.
developing nations despite its importance for public safety
improvement. This research presents a novel crime cluster-
ing model, CriClust, based on a dual threshold scheme for
crime series pattern (CSP) detection and mapping to derive
useful knowledge from a crime dataset. Based on analysis of
5500 (rape) crime records across 40 locations (suburbs) in
Western Cape, CriClust led to the identification of up to three
series at some of the locations investigated. We present an
effective web-based system that security agencies can use for
timely CSP identification to aid strategic and viable means
of combating crime in low resource settings. Fig. 1. A depiction of serial predator in related crime
scenarios in a city
Keywords— Public safety, Low-resource settings, Crime se-
ries
In promoting the identification of crime series, we present
1. INTRODUCTION CriClust, a model for revealing CSP information in crime
data. In this research, a user-centred system was developed to
South Africa (SA) is one of the countries with the highest elucidate how security agencies in low-resource settings can
crime rates around the world [1]. Therefore, the govern- be assisted in achieving tactical and strategic interventions.
ment’s vision is to invent new strategies for improved public A related study (by the authors) is focusing on how to inte-
safety outcome [2]. Tackling crime effectively is more chal- grate crowd-sourcing and mobile phones to promote crime
lenging in resource constrained settings, where crime intel- knowledge support. This can enhance operational planning
ligence experts and police are limited and not enough tech- with deep insights from crime data.
nological solutions are in place to meet up with daily op-
The rest of the paper is structured as follows: Section 2
erational safety needs. This is the case in most developing
provides a general background to current practices in South
nations where local police stations still adopt the traditional
Africa and presents a summary of gaps and opportunities that
(manual) means of crime data collection and by extension
have been identified in the current system and approach to
limited analysis for knowledge support, which hinders strate-
knowledge support for crime control. The key contribution
gic and tactical crime interventions. Moreover, this limited
of this paper is described in sections 3, 4, 5 in terms of the
data analysis for knowledge support does not align with the
methodology, results, the web-based system and contribu-
tion to smart city development. Finally, section 6 presents
The authors gratefully acknowledge the financial support from the Na-
tional Research Foundation (NRF), South Africa. the conclusion and possible extension to the research.
978-92-61-24291-6/CFP1768P-ART © 2017 ITU – 89 – Kaleidoscope