Extensible Data Management Architecture for Smart Campus

Applications

A Crowdsourcing based Solution

Attila Adamk

o and Lajos Koll

Department of Information Technology, Faculty of Informatics, University of Debrecen, Debrecen, Hungary

Keywords:

Smart Campus, Crowdsourcing, Participatory Sensing, Data Integration.

Abstract:

The technological advancements that have occurred during the past decade in various domains, including

sensors, wireless communications, location positioning technologies and the web, allow the collection of

a wide range of data. Possible sources of that data include intelligent devices (smartphones, tablets, etc.)

containing various sensors, Web pages, and social networking sites. Collected data are subject to analysis

(using data mining or pattern recognition approaches, for instance) and after processing new content might

be inferred. This is a value-added service that can itself be used as a data source. In this paper, we use our

University Campus as an example for establishing a data management architecture that integrates into a more

general, extensible publish/subscribe based model of crowdsourced applications.

1 INTRODUCTION

The recent rise in ubiquitous computing (ubicomp)

and supporting devices resulted in lots of new ap-

plications that are able to exploit the advantages of

this approach. According to (Luo, 2012), “ubicomp

is a post-desktop model of human-computer interac-

tion in which information processing has been thor-

oughly integrated into everyday objects and activi-

ties”. The collected data from the built-in sensors of

smartphones, tablets, phablets, and embedded devices

let the applications based on participatory sensing be-

come more and more popular and useful. Moreover,

in a real-life environment we need additional (user-

or application-generated) data sources in order to pro-

vide valuable services.

A Smart Campus is a good prospect of applying

participatory sensing combined with other data and

event sources and due to the number of people study-

ing, working or even living at a Campus and consid-

ering their commitment of using novel technologies

and applications, it is an ideal choice for developing

value-added services based on crowdsensing.

This paper is organized as follows. Section 2 de-

scribes the context of our research and explains the

reasons why we selected Smart Campus for our fo-

cus. Key challenges and our research goals are also

deﬁned here. Section 3 brieﬂy describes the related

work. The main contribution of the paper, namely,

the supporting architecture is discussed in Section 4

while Section 5 contains some concluding remarks.

2 SMART CAMPUS AS A PART

OF A SMART CITY

The context of this research is a project entitled Future

Internet Research, Services and Technology (FIRST

)

has been started in 2012 by University of Debrecen

and its consortial partners. The research objectives of

the project include to face the challenges posed by the

Internet, to carry on investigations covering basic the-

oretical questions, network modelling examinations,

the reassessing of network architecture, content man-

agement, and the creation of an intelligent application

platform.

The aims of one of its subprojects, Future Internet

for Smart City Applications, are threefold:

1. to establish an open, mobile crowdsensing appli-

cation platform for smart city applications,

2. to provide data management and knowledge dis-

covery solutions for them, and

3. to develop prototype applications based on the

http://ﬁrst.tamop422.unideb.hu/

About-the-project.html?language=en

226

Adamkó A. and Kollár L..

Extensible Data Management Architecture for Smart Campus Applications - A Crowdsourcing based Solution.

DOI: 10.5220/0004964802260232

In Proceedings of the 10th International Conference on Web Information Systems and Technologies (WEBIST-2014), pages 226-232

ISBN: 978-989-758-023-9

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

Figure 1: XMPP-based publish/subscribe architec-

ture (Szab

o and Farkas, 2013a).

common platform combining the data manage-

ment solutions with real-time analytics.

The open platform has already been estab-

lished (Szab

o and Farkas, 2013a; Szab

o and Farkas,

2013b; Szab

o et al., 2013). Figure 1 shows how

the model can directly be mapped onto the Extensi-

ble Messaging and Presence Protocol’s (XMPP) pub-

lish/subscribe (pubsub) service (Saint-Andre, 2011).

During our work to design data management in-

frastructure, we started to investigate the ﬁeld of

Smart Campuses instead of Smart Cities (fortunately,

this also ﬁtted into the overall project). The reason for

it was that there are lots of similarities between the

challenges of developing Smart Campus and Smart

City applications (the corresponding terms used by

Figure 2 and (Szab

o et al., 2013) are given inside

parentheses):

Lots of Potential Users (Consumers)

University of Debrecen has as much students

and employees as the number of inhabitants of

a medium-sized Hungarian city. These users are

consuming the provided services.

High Variety of Data Sources (Producers)

Information in various formats coming from mul-

tiple sources should be integrated in both cases

(e.g., timetable, academic calendar, information

on consultation dates and times in case of Smart

Campus, or energy consumption, trafﬁc info, etc.

for Smart Cities). Social media sites and geoloca-

tion sensors are examples of sources that are com-

mon in both domains.

Need for Value-added Services (Service Providers)

Service Providers give added value to the crowd-

collected raw data. Basically we can say that this

is why someone would like to use the applications

(e.g., for being notiﬁed about a room change or

ﬁnding a jam-free route between two points).

Figure 2: Crowdsourcing model based on a pubsub

scheme (Szab

o et al., 2013).

Examples of such value-added services include

but not limited to:

• get notiﬁed when one of my favourite bands

(based on my listening history from last.fm or

Google Play Music) will play in a concert venue

of the Campus,

• get notiﬁed when an event’s (class, concert, etc.)

date and time is modiﬁed,

• based on the location data provided by some of

my friends I can realize that they are going to have

a lunch so if I am in a hurry I can join them,

• when preparing for a consultation with an instruc-

tor, get notiﬁed if too many students plan to visit

the instructor’s ofﬁce hour at the same time, and

assist of reorganizing my schedule based on my

task list.

We have a much longer list describing various use

cases for Smart Campus applications but since the fo-

cus of this paper is not on the applications but the un-

derlying data management architecture, we omit the

details of use cases.

An advantage of building a prototype application

on the Smart Campus ﬁeld over a Smart City is that

young people’s (especially students’) willingness to

use the applications and provide valuable feedback on

them is much bigger than elders’ therefore it seemed

to be a good idea to test and evaluate the underlying

data management architecture. On the other hand,

participatory sensing can successfully be applied as

students actively use handheld devices and social me-

dia (presumably they are much more active than city

inhabitants are). Our aim is to build services that can

be used to infer some patterns regarding the operation

of the community, and the derived information can be

utilized as a part of a new service offered for the com-

munity after applying appropriate analytics.

Smart Campus applications should ﬁt into this

framework so they need to follow an XMPP-based

pubsub approach with crowdsourcing. However, in

order to achieve platform independence, we decided

to encapsulate services into Web services which allow

ExtensibleDataManagementArchitectureforSmartCampusApplications-ACrowdsourcingbasedSolution

227

to develop consumer apps on any of the most popu-

lar mobile platforms (Android, iOS, Windows Phone)

along with traditional Web applications.

2.1 Challenges

A Smart Campus environment has lots of various

(and most importantly, heterogeneous) data sources

including the following:

• an Education Administration System called Nep-

tun that contains information on course enroll-

ments, timetable information of courses, exam

dates and times, etc.,

• faculty members offering ofﬁce hours, consulta-

tions, etc.,

• Education Ofﬁces of the various faculties offering

ofﬁce hours,

• Student Governments organizing events for stu-

dents,

• the menus of the canteens located at the Campus,

• geolocation (e.g., GPS), WiFi or some other sen-

sor data collected by smartphones or similar de-

vices,

• data gathered by environmental and building sen-

sors (temperature, humidity, air pressure, air pol-

lution, etc.),

• a Library Information System that is able to tell

whether a given book is available or not,

• social media sites (like Facebook or Google+)

containing information on friends and ranges of

interests of a person,

• professional sites (like LinkedIn) holding data on

work experience and professional achievements

(however, this is not necessarily the most impor-

tant data source from Smart Campus perspective),

• bibliographic databases (like Google Scholar,

DBLP or Scopus) that provide information of

published journal articles or conference papers of

researchers,

• event hosts of actually any events (like public

lectures, concerts, exhibitions or whatever users

might be interested in), and, which is essential,

• the crowd itself with the added value of the capa-

bility of generating content that is interesting for a

set of people (or, to be more precise, consumers).

Of course, this is not an exhaustive list, these have

only been listed in order to demonstrate how diversi-

ﬁed sources of data can be used in a complex appli-

cation. Valuable applications typically require inte-

gration of some data originating from several of these

sources. Therefore, our main challenge is to create an

architecture which allows the access of information

from existing sources while also easing the addition

of new sources in a seamless manner.

Some of those data can be gathered in an auto-

mated way (like sensor data), some of them may re-

quire manual interaction (e.g., canteens’ menus or in-

structors’ ofﬁce hours); some of those data sources

offer Application Programming Interfaces (APIs) to

provide access to data (social media sites, for in-

stance) while others do not have APIs so some web

spiders are required to parse the data; some of the

data sources provide built-in notiﬁcation mechanisms

(e.g., an event feed of a social network site) while oth-

ers do not (for example, adding new ofﬁce hours or

changing the daily menu).

Considering the amount of data originating from

those sources, we can state that a big data manage-

ment solution is required. The frequency of arriv-

ing data elements can signiﬁcantly differ, some of

them (e.g., inserting a new article into a bibliographic

database or offer ofﬁce hours for the forthcoming

semester) might be quite rare while others (especially

when collecting location data) can be very frequent.

This will be further discussed in Section 4.

2.2 Our Goal

An architecture for data management and knowledge

discovery is needed that ﬁts into the framework shown

in Figure 1. This architecture should address the chal-

lenges described in Section 2.1 and be as generic as

it can be in order to integrate a high variety of data

sources regardless of how the data is processed. Ex-

tensibility is a key concern: the design should allow

to add new sources and deﬁne their processing in an

easy way.

Besides setting our goals, it might be useful to dis-

cuss what are the non-goals. On one hand, discus-

sions on the use of XMPP will not be provided as the

overall framework was given. On the other hand, the

development of value-added services may (and will)

require data mining activities (including understand-

ing the semantics of data in order to eliminate dupli-

cations), as well, however, they are completely out of

scope of this paper as we here focus only on the data

management platform that can (and should) later be

complemented with on-line data analytics solutions

which are currently under development (by some of

our colleagues). Data security is an important issue

that needs to be addressed in the technology stack,

however, our opinion is that it must be introduced at a

higher level than the data management level, therefore

we omit the discussions regarding that ﬁeld.

WEBIST2014-InternationalConferenceonWebInformationSystemsandTechnologies

228

3 RELATED WORK

There have been innovative steps in Smart City devel-

opment over the past few years. It became essential to

develop a platform that will aggregate all the available

(related) information and will orchestrate it in the best

way as possible, towards meeting the deﬁned goals.

3.1 Smart Campus

There is an even growing literature for Smart Cam-

puses which are related to data integration and

real-time processing of massive heterogeneous data

sources. An application prototype using semantic

technologies is discussed in (Boran et al., 2011). The

idea behind this approach is to include semantic in-

formation using ontologies with OWL and SPARQL.

Another approach (Valkanas and Gunopulos, 2013)

deals with event detection from real-time data using

heterogeneous sources.That experiment shows that

information extracted form social networks and sen-

sor data can be used for identifying events and they

show how affected users (that are physically close to

the event source) will be notiﬁed based on their actual

geolocation.

The high majority of related literature approaches

Smart Campuses from an Internet of Things (IoT)

point of view, having a focus on the physical cam-

pus infrastructure including sensor networks built into

buildings, physical entities and places (e.g., parking).

This is promising from the energy efﬁciency point of

view but basically provides the same challenges as

Smart City projects do. As we were looking for a

bit different kind of challenges, we decided to turn

to the social factors more than infrastructural. Be-

sides ﬁnding the need for integration of data coming

from sensor networks very important we think that

non-infrastructural aspects should play an equally im-

portant role in a life of a Campus. Therefore, people

living, studying and working on the Campus should

heavily be involved so we feel that exploitation of the

crowd’s power has a crucial effect on the success of

any Smart Campus-related projects.

3.2 Crowdsourcing

The term crowdsourcing is not easy to deﬁne. It is

a hybrid of “crowd” and “outsourcing”, meaning that

a given task is assigned to a number of people work-

ing on it explicitly. This approach is too restricting,

as it is coined in (Doan et al., 2011), therefore they

give the following deﬁnition: “a system is a crowd-

sourcing (CS) system if it enlists a crowd of humans

to help solve a problem deﬁned by the system owners,

and if in doing so, it addresses the following four fun-

damental challenges: How to recruit and retain users?

What contributions can users make? How to combine

user contributions to solve the target problem? How

to evaluate users and their contributions?”

Starting from these challenges, (Doan et al., 2011)

also gives a classiﬁcation of CS systems. Nine di-

mensions are identiﬁed but here we only concentrate

on those that are relevant for us. Based on the nature

of collaboration, CS systems can be either explicit or

implicit. Both are important for us as explicit systems

include evaluating (rate courses, meals, etc.), sharing

(location info, ﬁles, etc.) and networking (adding new

friends or classmates) while implicit systems might

be either standalone (when event prediction is based

on the history of users’ activities) or piggyback (for

example, based on the trajectories of a user’s move-

ment, predictions can be made). In the context of

Smart Campus applications, for the ﬁrst fundamental

challenge (recruiting users) only voluntary participa-

tion is possible. For more classes of CS systems with

detailed explanations, see (Doan et al., 2011).

3.3 Graph Databases

Graph databases are gaining popularity in the past few

years as they provide adjacency structures for data

elements without having any indices (i.e., data ele-

ments contain direct pointers showing their adjacent

elements). The three basic building blocks of graph

databases are nodes (representing entities), proper-

ties (pertinent information related to nodes) and edges

(connecting nodes to nodes using directed arcs). Ac-

cording to the Property Graph Model (Robinson et al.,

2013, p. 4), nodes contain properties that are key—

value pairs, and edges (a.k.a. relationships) can also

have properties. This organization of data allows

them to be processed using the well-known graph al-

gorithms.

3.3.1 Advantages of Graph Databases

According to (Robinson et al., 2013), graph databases

offer three key advantages:

Performance. Graph databases are easy to scale

since queries are localized to a portion of the

graph only therefore query execution time is pro-

portional only to the size of the affected sub-graph

rather than the size of the overall graph.

Flexibility. Graphs are “naturally additive, meaning

we can add new kinds of relationships, new nodes,

and new subgraphs to an existing structure with-

out disturbing existing queries and application

ExtensibleDataManagementArchitectureforSmartCampusApplications-ACrowdsourcingbasedSolution

229

Web Service

XMPP server

HistoricalDB

Data Access Layer

Connector

Smart Campus

Central Intelligence

Local LDAP

Local WWW

Local Educ. Sys.

. . .

Connector

G+

Facebook

...

DBLP

Scopus

G. Scholar

Connector

GPS

WiFi

...

Figure 3: Our extensible architecture.

functionality. These things have generally pos-

itive implications for developer productivity and

project risk. Because of the graph models ﬂex-

ibility, we do not have to model our domain in

exhaustive detail ahead of timea practice that is

all but foolhardy in the face of changing business

requirements.”

Agility. Since the graph data model has a schema-

free nature, graph databases suit well for iterative

incremental style software development practices.

Even the evolution of the data model can be per-

formed in small, agile steps which is a huge ad-

vantage when talking about data integration. One

can easily add new properties caused by emerging

data sources to the existing data model without

the need of performing complex schema evolu-

tion steps (which are, in fact, probably the biggest

challenge when using relational databases).

The graph data model ﬁts well to the data structure of

many of today’s popular social networking sites in-

luding Facebook, Twitter, or LinkedIn. It is a deﬁni-

tive advantage over relational databases, key–value

stores, and XML databases when our goal is to in-

tegrate data from such sources.

4 EXTENSIBLE DATA

MANAGEMENT

ARCHITECTURE

One of the primary objectives of our system is to cap-

ture continuous data streams arriving from different

sources. As we discussed in Section 1, Extensible

Messaging and Presence Protocol (XMPP) protocol,

deﬁning a standardized way of event-based messag-

ing, has been selected as the underlying communica-

tion protocol (Szab

o and Farkas, 2013b), due to its ex-

tensibility and publish/subscribe model. Based on the

heterogeneity of the data we should prepare proper

input channels for them.

Our proposed architecture (see Figure 3) intro-

duces the term Connector which is a node responsi-

ble for collecting data from given sources and send

updates to the Smart Campus. The notiﬁcation—

and data transfer—is achieved by sending a special

XMPP message to the Smart Campus Collector Inter-

face (as shown in Figure 4). Using this information it

can be decided whether to approve or decline any up-

dates based on the gathered information. The reposi-

tory update is declined when a Connector sends a re-

dundant piece of information, for instance, when the

Research Connector integrating several sources like

DBLP, Scopus, Google Scholar, etc. realizes that it

collects the same piece of data that has already been

found earlier.

The Connectors in our proposal are crowdsourced

compound entities without any deep business logic.

They containing the necessary parser or API imple-

mentation for the proper sources for a given ﬁeld—

like a parser for Google Scholar and an API for Sco-

pus. It can be imagined as an open box which can

be extended (for existing domains) or created and at-

tached to the system as a brand new one—as a new

and previously not existed ﬁeld. This is the power of

WEBIST2014-InternationalConferenceonWebInformationSystemsandTechnologies

230

Users

Services

Central

Intelligence

Data Access

Sources

XMPP

Models

Figure 4: Smart Campus layers.

the architecture because it can be done by the crowd.

The only requirement is to prepare the Connector for

the proper XMPP message structure. If the XMPP

server has not got the required message type the ex-

tension mechanism of the XMPP comes into the pic-

ture. It can be created, inserted and published to the

server.

Figure 4 shows a layered view of our Smart Cam-

pus architecture. The bottom layer shows the various

(and incomplete) sources that are used for gathering

data.

The Data Access Layer also plays an important

role in our architecture. It provides quick access to the

more-or-less statical personal data and to the “fresh”

real-time data. It means, the frequently changing data

(like sensor data, news feed) can be easily accessed

only for a limited time, e.g., three or six months and

after all it goes to the HistoricalDB part. This sep-

aration can speed up the queries targeted for “fresh”

data because the data store cannot grow to a huge one

slowing down all the searches. However, the other

part, the historical one introduces new challenges as

well. At one hand, efﬁcient handling of big data. On

the other hand, proper support for the Analytics mod-

ule to ﬁnd new and relevant information from that

huge amount of data. Since our original goal was to

design a data management platform, development of

the Analytics module has been postponed until a sub-

sequent iteration.

The Smart Campus Central Intelligence (SCCI)

component in our architecture provides an interface

between the information sources including both the

Applications

Central

Intelligence

Data

Management

Data

Collector

Conn

ectors

Models

Figure 5: Smart Campus services.

incoming events (XMPP server) and the information

stored in the database and the Web services layer (Fig-

ure 4). A service-oriented view of our architecture

is shown in Figure 5. The Data Collector layer uses

Connectors to connect sources for receiving data. The

collected data is then used to populate the databases

of the Data Management layer. Applications that are

created based on Web services are at the top of the

view.

The underlying data model for Smart Campus also

highlights challenges. We need to face structured and

unstructured data as well. Moreover, it needs to be an

extensible model which leads us to a semi-structured

model where we know it does not ﬁt into a uniform,

one-size-ﬁts-all, rigid relational schema. Faced with

the need to generate ever-greater insight and value-

added services we turned to graph technologies to

tackle the complexity. Besides increasing volume of

data, another force which has to be contended with

is that over time, data will become more and more

unstructured. As data volumes grow, we trade in-

sight for uniformity; the more data we gather about

a group of entities, the more that data is likely to be

semi-structured but this the way what we originally

imagine, an extensible system. But insight and end-

user value do not simply result from collecting large

volume and variation from data. Insight depends on

understanding the relationships between entities, so

we need to map how the entities in our domain are

connected. The key thing about such a model is that it

makes relations ﬁrst-class citizens of the data, rather

than treating them as only metadata. With this ap-

proach we established a data store which can be ex-

tended in an easy way. When a new—and previously

ExtensibleDataManagementArchitectureforSmartCampusApplications-ACrowdsourcingbasedSolution

231

not existed—source is attached to the system we only

need to add the new nodes and proper edges to the

database.

5 CONCLUSIONS

In this paper we brieﬂy described a data manage-

ment architecture for Smart Campus applications.

This architecture ﬁts well into the more general pub-

lish/subscribe based architecture of Smart City and

Smart Campus applications. The developed archi-

tecture is extensible with new data sources (appro-

priate Connectors need to be developed when adding

a new source) providing the capability of integration

of heterogeneous data. Our future plans include to

develop more applications based on that architecture

along with the addition of the Analytics module.

ACKNOWLEDGEMENTS

The publication was supported by the T

AMOP-

4.2.2.C-11/1/KONV-2012-0001 project. The project

has been supported by the European Union, co-

ﬁnanced by the European Social Fund.

REFERENCES

Boran, A., Bedini, I., Matheus, C. J., Patel-Schneider, P. F.,

and Keeney, J. (2011). A smart campus prototype for

demonstrating the semantic integration of heteroge-

neous data. In Rudolph, S. and Gutierrez, C., editors,

RR, volume 6902 of Lecture Notes in Computer Sci-

ence, pages 238–243. Springer.

Doan, A., Ramakrishnan, R., and Halevy, A. Y. (2011).

Crowdsourcing systems on the world-wide web. Com-

mun. ACM, 54(4):86–96.

Luo, X. (2012). Social network-based media sharing in

the ubiquitous environment: Technologies and appli-

cations. In Abraham, A. and Hassanien, A.-E., edi-

tors, Computational Social Networks, pages 349–366.

Springer London.

Robinson, I., Webber, J., and Eifrem, E. (2013). Graph

Databases. O’Reilly Media.

Saint-Andre, P. (2011). RFC 6120: Extensible Messaging

and Presence Protocol (XMPP): Core.

Szab

o, R., Farkas, K., Isp

any, M., Bencz

ur, A. A., B

atfai,

N., Jeszenszky, P., Laki, S., V

agner, A., Koll

ar, L.,

Sidl

o, C., Besenczi, R., Smajda, M., K

er, G., Sz-

incs

ak, T., K

adek, T., K

osa, M., Adamk

o, A., Lend

ak,

I., Wiandt, B., Tom

as, T., Nagy,

A., and Feh

er, G.

(2013). Framework for smart city applications based

on participatory sensing. In Proceedings of the 4

IEEE International Conference on Cognitive Infocom-

munications. to appear.

Szab

o, R. L. and Farkas, K. (2013a). A publish-subscribe

scheme based open architecture for crowd-sourcing.

In Bauschert, T., editor, EUNICE, volume 8115 of

Lecture Notes in Computer Science, pages 287–291.

Springer.

Szab

o, R. L. and Farkas, K. (2013b). Publish/subscribe

communication for crowd-sourcing based smart city

applications. In Proceedings in Conference of Infor-

matics and Management Sciences, pages 314–318.

Valkanas, G. and Gunopulos, D. (2013). Event detec-

tion from social media data. IEEE Data Eng. Bull.,

36(3):51–58.

WEBIST2014-InternationalConferenceonWebInformationSystemsandTechnologies

232