You are on page 1of 7

Application of Social Network Analysis to Port

Community Systems
Saša Aksentijević 1, Dražen Marković 2, Edvard Tijan 3, Mladen Jardas 4
1
Aksentijević Forensics and Consulting, Ltd.
Gornji Sroki 125a, Viškovo, Croatia
Tel: +385 51 65 17 00 Fax: +385 51 65 17 81 E-mail: axy@vip.hr
2
Neos, Ltd.
Budmanijeva 1, 10000 Zagreb, Croatia
Tel: +385 1 5555 600 Fax: +385 1 5555 620 E-mail: info@neos.hr
3, 4
University of Rijeka, Faculty of Maritime Studies
Studentska 2, 51000 Rijeka, Croatia
Tel: +385 51 33 84 11 Fax: +385 51 33 67 55 E-mail: etijan@pfri.hr, mjardas@pfri.hr

Abstract – Social Network Analysis is a unique and “Croatian Agency for SMEs, Innovations and
innovative concept in the global business management Investments” (HAMAG-BICRO) was founded with the
market that allows prediction of behaviour of various aim of supporting the development of small and
involved nodes (typically clients or customers) in a medium-sized enterprises, improving the innovation
structured dependent network using commonly provided
process and encouraging investments [3].
services. This methodology can be used by
telecommunication providers to offer new targeted
products before competition, prevent customer migration Government of the Republic of Croatia is using a
to the competition and rapid response to disruptive loan of the International Bank for Reconstruction and
technological and economic changes in the market. There Development for execution of the Second Croatia
are a number of solutions supporting telecommunication Science & Technology Project – STPII that was started
providers in provision of SNA-based solutions and at July 31st 2013 and will be in effect until March 31st
campaigns and the methodology can be adapted also for 2019. After 2015/2016 the tenders are closed and
banking and finance industry. SNA offers handling of projects are in execution phase. Main goal of the STPII
large calculations with adjustable parameters, project is to enhance absorption of EU funds in the area
adaptability of calculations according to business
of research and development by strengthening
requirements and application to diverse IT
infrastructures. In this paper the authors will present the capacities of users and institutions [4].
main architecture of a viable SNA system and propose a
novel application to the area of port operations with HAMAG-BICRO is (as a loan beneficiary) in
integration into existing Port Community Systems. charge of executing five various programs of support to
public and private users in various phases of the
Keywords: Social Network Analysis; Port Community innovation process:
Systems; ports; Big Data
1. PoC (Proof of Concept),
I. INTRODUCTION 2. RAZUM (Development of knowledge-based
Social network analysis (SNA) is not a formal enterprises),
theory, but in fact a strategy used to investigate various
social structures [1]. While the roots of SNA are 3. TEHCRO (Development of technology
derived from sociology, this method has found a infrastructure),
widespread use in other practical applications, and
4. IRCRO (Cooperative research and
especially in data aggregation and data mining,
modelling of networks, user and client behaviour and development), and
churn calculations and various other business 5. UTT (Support to Offices for technology
intelligence needs. SNA can be used in a variety of
transfer).
use-cases where appropriate dataset is available, so, for
example, it is also used in counter-terrorism operations, Within the framework of IRCRO program, in May
especially in calculation of centrality measurement as 2015 project documentation has been submitted to
an investigative tool [2]. HAMAG-BICRO according to program propositions,
by Neos Ltd. from Zagreb, Croatia, titled “Neos SNA
for Telco – an innovative system for anticipation of
This work has been financially supported by University of
mobile telephony users” [5]. Project partner was the
Rijeka under the Faculty of Maritime Studies projects.

MIPRO 2018/DE-GLGPS 1507


Faculty of Electrical Engineering and Computing from When SNA methodology is used, the system
University of Zagreb. Anticipated duration of the automatically uses consolidated data contained in data-
project was 18 months and final deliverable is a warehouse systems. Only necessary and strictly
complete solution for SNA analysis of mobile required data set is automatically loaded to SNA engine
telephony users, applicable also in other similar (consolidated data, data from transaction systems like
scenarios. CRM and billing and CDR – Call Data Record) data
that is generated directly in mobile exchange and
The project started in January 2016 and it was contains all relevant data for various communications.
successfully completed within budget, time and quality SNA enables employees of the marketing and sales
by the end of June 2017. Project manager on behalf of departments to run generation of SNA network model
Neos, Ltd. was Mr. Dražen Marković, project manager and various use-case analysis, with custom parameter
of the Faculty team was Ms. Bojana Dalbelo Bašić, adjustment.
PhD, while technical business project consultant acting
on behalf of HAMAG-BICRO was Mr. Saša Taking into consideration authors’ extensive
Aksentijević, PhD. experience with development of comparable systems
using traditional approach, direct advantages of SNA
During the evaluation of the end result of the methodology over traditional approaches are the
project, it became clear that developed software following:
package can also be used for other purposes where
structured data is available. Considering author's 1. Input data is prepared automatically, and does
interest in Port Community Systems (PCS), one not involve operator work,
possible path of further development of SNA
methodology will be proposed in this paper after 2. Analysis is run independently from IT
project results are presented. departments, there is no need to rely on
availability of the other department and
analysis can be run whenever it is required,
II. SOCIAL NETWORK ANALYSIS METHODOLOGY and as often as necessary,
Traditional analysis of users’ behaviour is
increasingly difficult to execute because of the 3. Data manipulation is executed using modern
following reasons: Web application, where operator access can
be determined according to internal
1. Analysed sample is too small, information security rules and practices,

2. Number of set parameters is very limited, 4. Analysis is controlled via parameters, and
there is no manual adjustment and long batch
3. Users are grouped very coarsely and the same processing,
rules are applied to all users in a group.
5. Entire available data set is used, not only a
This approach to analysis and churn prevention small, often unreliable sample,
(maintenance of the existing user base) requires
consolidation of data in data-warehouse systems of 6. Results are independent from the knowledge
mobile telephony operators, additional data aggregation and skill of operators,
that is usually done manually, this data is then exported
to Excel format used by marketing and additional 7. Finally, and most importantly, SNA
application of contextual analysis. methodology does not analyse only one
subscriber, but also all of his/her connections,
In this scenario the analysed data is prepared hence the name of the methodology.
manually, which takes a lot of time and work of IT
specialists, so it is generally not possible to satisfy on-
III. NEOS SNA FOR TELCO PROJECT BASICS
demand requests. Consequentially, marketing and sales
departments depend on others to do their job. Innovation of the SNA for Telco project is in
Considering the volume of data, only small subset is integration of SNA methodology, applied mathematical
analysed and this subset does not have to be theories and big data technology that enable quick and
representative for the whole set. Finally, this type of efficient data processing and calculations of required
analysis relies on properties of a single subscriber analysis for exact business activities and timely
(service user) and its attributes, like age, sex, address, decision making.
used services, consumption average etc. However, in
reality, all these attributes are less important than Therefore, this product not only enables new
context and user’s relation to other users (subscribers), analytical-predictive possibilities, but also a number of
because in social context, behaviour of a single user other advantages over classical approach like work
depends on his/her relation to other users. automation, lack of IT department involvement in
processes of the marketing and sales, more comfort in

1508 MIPRO 2018/DE-GLGPS


work and additional features for all stakeholders.
Finally, usage of SNA methodology can enable
decision makers to keep their customers, enhance
product packages and attract new customers. The
following Figure shows the visualisation of SNA
model with a small number of objects.

Figure 3. SNA system technological components

Development project itself consisted of several


stages.

1. Stage 1 – Preparation of the project plan. In


this phase, project application was prepared,
detailed task and assignment roster was
created, partner contracts were prepared,
financing application was drafted and project
plan was agreed between all stakeholders.

2. Stage 2 – Preparation of algorithms to set up


SNA network and individual analytical
algorithms. In this stage, various algorithms
Figure 1. SNA network that create general SNA network were created.
They were prepared for various Use-Cases:
SNA for Telco has several advantages over existing
products. It can be installed in-house, using  Churn Prevention
infrastructure of the operators, instead of moving data o Probability calculation & Detection
across borders, that can cause problems related to of churn-prone customers
privacy and data management, even if the data is o Leaders / Influencers identification
anonymized. Also, this product is highly customizable o Churn Prevention Lift Analysis
with adjustable parameters. All calculations can be  Campaign Management (SNA based)
adjusted to real work case scenarios of operators, they o Early Adopters detection
can be localized, new modules can be added and it can o Spinners identification
be installed across various infrastructure. General o Retention campaigns
schematic depiction of the system is shown in Figure 2. o Viral Marketing & Acquisition
System is fed with general existing data (CDR files, Campaigns
data from CRM and billing systems), SNA model is o Value segmentation (Low ARPU
prepared and analysis and export is prepared according subscribers with high RPU
to various campaign scenarios. connections)
 Household Identification – Family
Identification
o Customer segmentation
o Location / Time of call Detection

3. Stage 3 – Development of the prototype. In


this stage, a platform for development of the
complex solution was set up. Architectural
components of the solution and possibilities of
their interconnection were defined. Finally,
Figure 2. SNA system component blueprint development system was set up. Some parts of
the system were functioning in a virtual
Technological components of the system are based environment (Big Data Cloudera servers,
on various Oracle technologies where each technology Oracle database with Oracle Advanced
is used for particular function. Layout of the Analytics), and application server for GUI.
underlying technology is presented in Figure 3. This prototype was also defined and

MIPRO 2018/DE-GLGPS 1509


documented according to best practices and and its modules. Algorithms were prepared based on
set up for further development activities. mathematical theories, which require high level of
quantitative knowledge. This particular risk was
4. Stage 4 – Check of the quality and the mitigated by the inclusion of the research team that
functioning of the algorithms. In this phase, possessed significant experience in the above
prototypes were integrated with the database, mentioned field, especially in the field of mathematical
and connected with parameter entry tables. theories.
Error checking was also completed. Various
calculations were checked (depending on the The second groups of identified risks were market
parameter set) against the expected results risks. Namely, it was possible that operators are not
based on algorithms. End results was that the adequately interested in advanced analytical solutions.
best parameter sets were selected for SNA Before the project, the level of innovation and market
network and various Use-Case calculations interest was tested. The level of innovation was
(parameters giving the best results are acknowledged by winning the “Oracle Innovation
detected). Award for Big Data & Analytics” on the level of
Eastern Central Europe [6]. Furthermore, some initial
5. Stage 5 – Set up of the test infrastructure contacts with several mobile telephony operators, who
similar to production environment. In this expressed their interest in the above mentioned
stage, three main groups of activities were solution, were made.
performed. Firstly, integration services were
developed, with end result being the defined All anticipated risks were successfully managed
infrastructure base and all solution during the project and did not have a negative impact
components similar to quantity and capacity on the final results.
of data processing to the production
environment. Module set for data input, A set of standard project software documentation
transfer and transformation to a form suitable was created during the project execution that
for calculations and export was also comprised of the following components:
completed. Secondly, GUI (Graphical User
Interface) for visualization and data 1. SNA System Technical Architecture,
management was developed. This module is
used to integrate parameter setting, data input 2. SNA System Algorithms,
and export, log analysis and user and rights
3. SNA System Description,
management into a single console.
4. SNA System UI Description, and
6. Stage 6 – Generation of large data sets
similar to production data. As a part of this 5. SNA - Final Presentation.
activity, program modules that are not a part
of the solution (external utilities) were Oracle BDA is cluster of servers as a standard
developed and used to generate CDR files, but configuration, and is used in identical manner as in a
also other data necessary for solution testing. prototype environment.
These utilities have to closely emulate logical
data sets used by the mobile operators. Oracle Database is clustered with common storage
infrastructure. Real Application Clusters option (RAC)
7. Stage 7 – Testing of large data set processing is used for smart cluster environment infrastructure.
and system adjustment. In the final project Specific options can be used if needed.
phase, model parameters for SNA network
and various Use-Case analysis were tested and Clustered application server environment is
optimized, depending on the processing speed, recommended for load balancing, high-availability
amount of processed data and expected application environment. Apache Tomcat or Oracle
results. Also, load, stress, integration and Glassfish servers are clustered with common http
analytic tests were performed. server in front.
IV. NEOS SNA FOR TELCO PROJECT RESULTS Oracle Linux 6 is used as an operating system for all
During the project preparation, several project risks mentioned servers.
were identified. During the project execution they were
successfully managed. Recommendation for production infrastructure was
to use Oracle BDA, as it was used in the product
The first possible group of identified risks were prototype. But to ensure scalability and robustness of
technological risks. SNA uses a number of connected the production environment, it was advisable to use
platforms, used as infrastructural solutions. This risk clustering environment. This clustering environment is
was managed by the selection of best-of-breed platform shown in detail in the following Figure 4.

1510 MIPRO 2018/DE-GLGPS


Figure 5. Neos SNA big data security [9]

Cluster was fully integrated with LDAP, so it does


not use any local users, thus further increasing the
simplicity of the security management. Kerberos
authentication is used along with HUE SSO integration
that is also LDAP integrated.
Figure 4. NEOS SNA for Telco Product Infrastructure [7]

In the final product, CDR generator was created as a V. PREREQUISITES FOR SNA INTRODUCTION IN PCS
module of the SNA application. Its generation was While being actively involved in development of
based on templates, meaning that it is parameterized described SNA project for mobile telephony operators,
and users are able to customize output. Files are stored the authors came to realization that similar SNA engine
on the Glass Server and they can be automatically might be modified to be deployed in Port Community
transferred to the BDA Landing Area. Systems, for two separate proposed purposes:
SNA visualization is a tool of the overall SNA 1. The Port authority could utilize SNA to
application framework that provides insight into Node analyse connections between various other
Ego Network. It combines communication data with stakeholders involved in cargo attraction to a
SNA metric results available in the table to the system certain port. There is a recent paper dealing
operator. SNA visualization can be altered to show with similar topic, but it used SNA approach
adjustable levels of display (1-3). to analyse research trends related to ports.
This paper is in fact a cross-sectional
Calculation repository contains a set of database longitudinal study of various other researches
objects (tables, views, sequences and triggers). It was in period between 1985 and 2015. A number
designed to facilitate calculations and provide of studies and contained keywords were
connections between input data, input parameters and analysed with main goal being to obtain
results. RStudio Server was used as integrated degree centrality and betweenness centrality
development environment (IDE) for R development. that would point to identification of research
Access is achieved through web browser and its trends related to port competition [10]. This is
implementation achieved movement of the in fact a form of meta-research.
computation closer to the data. R code is executed
directly from the source editor. 2. Customs authorities and ministry of interior
(Maritime Border Police) could use SNA to
A significant effort was undertaken in order to analyse connections between various
ensure big data security, which is of utmost importance stakeholders using documents and data
to mobile operators, and especially with the emergence available to Customs and Police during cargo
of the new GDPR regulation [8]. HDFS security was processing. One prerequisite for full-scale
used in order to ensure security on the folder level, implementation of such system is the
while Sentry was used for the Hive security. Various deployment of e-Manifest information system
layers of big data security are shown in Figure 5 [9]. [11]. Considering that in early 2016 the
European Maritime Safety Agency, together
with some other agencies and bodies has
launched the e-Manifest pilot project,

MIPRO 2018/DE-GLGPS 1511


preliminary data availability should not be a message creation and create a workable set of
significant issue. production data.

In both cases, it is important to understand that 5. Development of configuration and parameter


overemphasis on nodes tends to disregard "space" repository used for calculations, validations
between the nodes, and in this case, ports. By analysing and storage of system-specific statistical data.
only ports and cargo being trafficked between ports,
underlying connections are being largely disregarded. 6. Programming the module used to import data
However, ports in a network cannot be separated from into the system (Big Data layer)
the very fabric that creates such networks like a mesh
of other players in the logistics chain: railway and truck 7. Creation of background-ran modules with
terminals or even airports. So, utilization of SNA various calculations that process data in the
analysis imposed over the port community should be Big Data system (for example, Hive), while
carefully constructed in order not to give too big of an parameters for calculations and validations are
emphasis to certain nodes, and completely disregard stored in a separate repository in a relational
how those nodes "came to being". database that is also used for storage of static
data related to every individual calculation
and validation process.
VI. PROPOSAL OF APPLICATION OF SNA TO PCS
In case of mobile telephony operators, they are 8. Envisaging and creating a module used for
usually data owners so it is reasonably simple to gain graphical visualization of the calculation
access to data and create SNA models. However, in the (applicable to SNA network and connections,
case of PCS, the situation is much more diverse and and calculated metrics).
more complicated. While the PCS might contain all
Entire process is finalized by creation of the module
necessary data that could be entered into the SNA
for end result export. This phase can be especially
model, the PCS itself is formed around commercial
management of the cargo and administrative ship important for a large or complex PCS with large traffic
management during its arrival, stay and departure from volumes, because data export is typically more
important for them than data visualization, which is
the port. Therefore, PCS will inevitably have to be at a
significant level of development in order to aggregate typically used as just an auxiliary tool.
all data from commercial operators in the port and their
Terminal Operating Systems (TOS). VII. CONCLUSION

The authors are proposing the following path for After 1950s, research of networks became a separate
adoption of SNA in PCS: topic of interest in sociology, but also in financial
theory and in practice – banking. In the last decades
1. Definition and research of exact technologies with the rise of social networks, it became a topic of
required to develop the project. While the interest in the field of technology and communications.
technologies are already tested and proven One such project was developed in a period between
during the development of SNA for mobile 2015 and 2017 in Croatia, and was funded by Croatian
operators, the level of complexity of cargo Agency for SMEs, Innovations and Investments. It was
logistics might require alterations of the developed in 18 months by the company Neos Ltd.
existing technological setup. from Zagreb, Croatia and the Faculty of Electrical
Engineering and Computing from the University of
2. Development of SNA for PCS prototype, Zagreb, Croatia.
including infrastructure (equipment, operation
systems, Big Data layer, additional modules), This particular SNA application is adjusted for use
development tools and background services, by mobile telephony operators who can create custom
calculation, user interface modules etc. analysis of the networks comprised of their users and
subscribers, calculate the network, identify relative
3. Creation of development infrastructure similar importance of network members and their impact to
to production infrastructure. other members. It can also be easily adjusted for usage
by banks and financial institutions in order to evaluate
4. Setting up modules that will be used to networks created by their customers, but also to
generate input data (similar to CDR evaluate how risky are some customers spending habits
generators) for PCS. Considering that PCS are and other financial habits. SNA application was based
systems based around exchange of messages mostly on Oracle and some other enterprise-level open
between various stakeholders, and that XML source technologies.
schemes should be readily available, it should
not pose a problem to reverse engineer the Considering that maritime cargo logistics is also
effectively a network created by the main stakeholders,

1512 MIPRO 2018/DE-GLGPS


the authors propose the implementation of SNA also in [2] Mullins, S “Social network analysis and counter-terrorism:
measures of centrality as an investigative tool”, Journal of
conjunction with the PCS information systems. Main Behavioral Sciences of Terrorism and Political Aggression,
steps to be undertaken to adjust SNA methodology for Volume 5, 2013 – Issue 2, p. 115-136
usage in PCS are identified, along with two most viable [3] Hamag-Bicro “About us”, http://www.investcroatia.hr/about-
purposes: risk assessment of customs office and police, us/ (accessed 24th Dec. 2017.)
and connection between remote and domicile ports, [4] Hamag-Bicro, “Bespovratne potpore – Programi podrške
which can be used to evaluate the possibilities of inovacijskom procesu“, http://hamagbicro.hr/bespovratne-
potpore/programi-podrske-inovacijskom-procesu/ (accessed
additional cargo attraction, thus increasing the 24th Dec. 2017.)
competitiveness of seaports and the whole trade route. [5] Neos Ltd. „Business plan – Neos SNA for Telco“, Zagreb, May
28th 2015
Further research of this topic, at least in Republic of [6] http://www.neos.hr/neos-wins-oracle-innovation-award-for-big-
Croatia, relies on impeding implementation of national data-analytics-in-ece-region/ (accessed December 24th 2017)
PCS that will commence mid-2018., and introduction [7] Neos Ltd. „SNA for Telco – SNA System Technical
of NSW (National Single Window) for maritime Architecture“, technical documentation v1.3, March 31st 2017,
p. 25
traffic.
[8] The EU General Data Protection Website,
https://www.eugdpr.org (accessed December 24th 2017
[9] Neos Ltd. „IRCRO SNA for Telco“
REFERENCES [10] Jeon Jun Woo, Wang Ying, TaeYeo Gi “SNA Approach for
Analyzing the Research Trend of International Port
[1] Otte, E., Rousseau, R. “Social network analysis: a powerful Competition”, The Asian Journal of Shipping and Logistics,
strategy, also for the information sciences”, Journal of Volume 32, Issue 3, September 2016, p. 165-172
information science. 28 (6), 2002, p. 441-453
[11] http://www.farrow.com/article-what-is-emanifest (accessed
December 24th 2017)

MIPRO 2018/DE-GLGPS 1513

You might also like