Towards Predictive Policing: Knowledge-based Monitoring of Social Networks

Abstract

Increasing the resilience of the society against disorders, such as disasters, attacks or threatening groups, is a major challenge. Recent events highlight the importance of a resilient society and steps which are required to be taken in resilience engineering. textit{A priori} the optimal way to handle such adverse events is to prevent them, or at least provide appropriate courses of preparation. The essential requirement for every kind of preparation is information about relevant upcoming events. Such information can be gained for example from social networks and can form the basis for a long-term and short-term strategic planning by security forces. For that purpose, we here propose an application framework for knowledge-based social network monitoring, which aims at predicting short-term activities, as well as the long-term development of potentially dangerous groups. In this work, a theoretical outline of this approach is given and discussed.

Full text

Towards Predictive Policing: Knowledge-based
Monitoring of Social Networks
Michael Spranger, Florian Heinke, Steffen Grunert and Dirk Labudde
University of Applied Sciences Mittweida
Mittweida, Germany
Email: {name.surname}@hs-mittweida.de
Abstract—Increasing the resilience of the society against dis-
orders, such as disasters, attacks or threatening groups, is a
major challenge. Recent events highlight the importance of a
resilient society and steps which are required to be taken in
resilience engineering. A priori the optimal way to handle such
adverse events is to prevent them, or at least provide appropriate
courses of preparation. The essential requirement for every kind
of preparation is information about relevant upcoming events.
Such information can be gained for example from social networks
and can form the basis for a long-term and short-term strategic
planning by security forces. For that purpose, we here propose
an application framework for knowledge-based social network
monitoring, which aims at predicting short-term activities, as well
as the long-term development of potentially dangerous groups.
In this work, a theoretical outline of this approach is given and
discussed.
Keywords–forensic; text processing; resilience engineering
I. INTRODUCTION
The representation and the communication via the Internet,
especially in social networks, have become a standard not
only for individuals, companies and organizations but also for
political groups or gangs using these platforms for planning,
appointing and conducting criminal offences [1], [2]. Large
events with a relatively large degree of group dynamics,
like sport events, demonstrations or festivals, require a high
expenditure of staff on the side of the security forces because
of unpredictability and uncertainty of associated dynamics. For
example, to secure the soccer events in 2014 in Germany
approximately two million working hours of police officers
were necessary [3]. In order to support decision makers, we
outline an application framework for monitoring cliques and
groups in social networks, which can be key elements in
the emergence of critical events. The monitoring process is
facilitated by means of employing general domain-specific
endangerer profiles. Such a profile can be deduced from a set
of social network sites of known endangerers or perpetrators
(in the strict sense). Identifying suspicious activities is realized
by group recommendation classifiers.
The following section is structured according to the steps
required to generate the proposed framework. First, aspects of
ontology definition are outlined, followed by discussions on
endangerer profile generation and classifier training. Finally,
monitoring strategies are proposed.
II. PRO PO SA L
The proposed application framework enables decision mak-
ers of security forces to identify threat hot-spots. In this way,
they are able to control their human resources. In order to
support long-term resource planning, The second aim is to
predict the long-term development of groups that pose a threat.
The process pipeline consists of three parts:
1) modelling the threat ontology
2) train the general domain-specific endangerers profile
3) monitoring all matching social network sites and
calculate a long-term and short-term threat score
A. Threat Ontology
The term ontology in a common understanding means a
formal and explicit specification of a common conceptualiza-
tion. In particular, it is defined as a set of common classified
terms and symbols referred to a syntax, and a network of asso-
ciate relations [4]. Similar to the crime ontology we proposed
in recent work [5], an ontology can be used for modelling a
complex threat assessment. In this way, knowledge of decision
makers is introduced and can be used for extracting semantic
information from posts and comments of social network’s
profiles. In particular, the works of Wimalasuriya and Dou [6],
Embley [7] and Maedche [8], show that the use of ontologies
is suitable for assisting the extraction of semantic units, as well
as their visualization and structures such processes very well.
B. Endangerer Profile
In order to distinguish profiles of interest regarding to
a certain threat, a general profile needs to be modelled.
Recent work [9], [10] has shown that feature vectors derived
from social network profiles are suitable for generating group
recommendations. In a similar way, a general classifier can be
trained based on the social network profiles of known persons
associated with a special threat. For example, Facebook pro-
files of known hooligans of a specific soccer club can be used
to train classifiers that are able to identify social activity of
hooligans and peers in social networks.
The generation process is divided into three parts depicted in
Figure 1.
C. Monitoring Activities
Once a profile is generated and the threat specific ontology
is defined, the social network monitoring can be conducted. At
this point a multi-level, information extraction process aims
at instantiating the ontology using textual information, like
posts and comments. An example of how such a process can
be structured is given by Spranger and Labudde [5]. Further
text analysis steps, like sentiment analysis (see the discussions
39Copyright (c) IARIA, 2015. ISBN: 978-1-61208-415-2
IMMM 2015 : The Fifth International Conference on Advances in Information Mining and Management

Profile
Selection
Social
Feature
Selection
Classifier
Training
Figure 1. The process of deriving a threat specific general profile.
given in [11] and [12] for details) can complete the instantiated
model in different ways. As a short-term benefit, a score can
be computed for various time points, signalling whether a
threatening event regarding to the specific profile and ontology
is directly pointing to a specific location and time frame. These
results can be applied to a map to localize short-term hot-spots
in terms of security and their dynamics as discussed by Davies
and Bishop [13].
Figure 2. The proposed system. The central, expert-modelled threat-specific
ontology describes the environment of a special threat. A general endangerer
profile completes the model. In the process the model is used to extract textual
information from social network activities. Different scoring functions allow
the identification of threat hot-spots or can show the long-term evolution of
groups and cliques.
In the age of Big Data and algorithms handling such
amounts of information, deducing long term developments of
such groups and dynamics is at its early stage. Methodological
concepts widely used in modelling complex relations (as for
instance systems biology) can be directly transferred to the
field of resilience engineering. Especially, employing generic
mathematical models to social networks has become compu-
tationally feasible, but requires further research. For example,
epidemiological models can be efficiently applied to study long
term evolutions of groups and sub-networks (see [14]) and
study the information transfer between them. Thus, generating
valid models and derive predictions from them can be of great
value, for instance, in planning personnel and staff demands.
REFERENCES
[1] ITU. Number of worldwide internet users from
2000 to 2014 (in millions). statista. [Online]. Avail-
able: http://www.statista.com/statistics/273018/number-of-internet-
users-worldwide/ (2015)
[2] eMarketer & American Marketing Association. Number of social
network users worldwide from 2010 to 2018 (in billions). statista.
[Online]. Available: http://www.statista.com/statistics/278414/number-
of-worldwide-social-network-users (2015)
[3] ZIS. Jahresbericht 2013/14. Zentrale Informationsstelle
Sporteins¨
atze. [Online]. Available: http://www.polizei-nrw.de/
media/Dokumente/Behoerden/LZPD/ZIS Jahresbericht 2013 14.pdf
(2014)
[4] T. R. Gruber, “Toward principles for the design of ontologies used for
knowledge sharing,” in Formal Ontology in Conceptual Analysis and
Knowledge Representation, N. Guarino and R. Poli, Eds. Kluwer
Academic Publishers, 1993.
[5] M. Spranger and D. Labudde, “Towards establishing an expert system
for forensic text analysis,” International Journal on Advances in Intel-
ligent Systems, vol. 7, no. 1/2, 2014, pp. 247–256.
[6] D. C. Wimalasuriya and D. Dou, “Ontology-based information extrac-
tion: An introduction and a survey of current approaches,” Journal of
Information Science, vol. 36, no. 3, 2010, pp. 306–323.
[7] D. W. Embley, “Toward semantic understanding: an approach based
on information extraction ontologies,” in Proceedings of the 15th Aus-
tralasian database conference - Volume 27, ser. ADC ’04. Darlinghurst,
Australia, Australia: Australian Computer Society, Inc., 2004, pp. 3–12.
[8] A. Maedche, G. Neumann, and S. Staab, “Bootstrapping an ontology-
based information extraction system,” Studies In Fuzziness And Soft
Computing, vol. 111, 2003, pp. 345–362.
[9] M. Manca, L. Boratto, and S. Carta, “Producing friend recommenda-
tions in a social bookmarking system by mining users content,” in Proc.
3rd. International Conference on Advances in Information Mining and
Management, IARIA. ThinkMind Library, 2013, p. 59 to 64.
[10] M. Cheung and J. She, “Bag-of-features tagging approach for a better
recommendation with social big data,” in Proc. 4th. International Con-
ference on Advances in Information Mining and Management, IARIA.
ThinkMind Library, 2014, p. 83 to 88.
[11] S. M. Mohammad, S. Kiritchenko, and X. Zhu, “Nrc-canada: Building
the state-of-the-art in sentiment analysis of tweets,” in Proceedings of
the Second Joint Conference on Lexical and Computational Semantics
(SEMSTAR’13), 2013.
[12] X. Wan, “Co-training for cross-lingual sentiment classification,” in
Proceedings of the Joint Conference of the 47th Annual Meeting of the
ACL and the 4th International Joint Conference on Natural Language
Processing of the AFNLP: Volume 1. Association for Computational
Linguistics, 2009, pp. 235–243.
[13] T. Davies and S. Bishop, “Modelling patterns of burglary on street
networks,” Crime Science, vol. 2, no. 1, 2013, p. 10.
[14] J. Cannarella and J. A. Spechler, “Epidemiological modeling of online
social network dynamics,” CoRR, vol. abs/1401.4208, 2014.
40Copyright (c) IARIA, 2015. ISBN: 978-1-61208-415-2
IMMM 2015 : The Fifth International Conference on Advances in Information Mining and Management