You are on page 1of 20

Towards the Implementation of the Learning Analytics in

the Social Learning Environments for the TechnologyEnhanced Assessment


Emilio Labra, Mara E. Alva, Juan M. Cueva
(University of Oviedo, Asturias, Spain
labra@lsi.uniovi.es, alvamaria@uniovi.es, cueva@uniovi.es)

Hernan Sagastegui Chigne


(Antenor Orrego Private University, Trujillo, Per.
tsagasteguic@upao.edu.pe)
Abstract: In this work is presented the focus formal of a framework of learning environment
that gives support to the learning analytics, based on the techniques that the educational data
mining and social network analysis provides. The purpose is to study or discover collaborative
relationships that students generated in their learning, make prediction or assessment of student
learning. Also is presented an alternative model to the model traditional in the form of
evaluating the learning. This model is more personal and to contribute to new possibilities that
digital technologies offer in technology-enhanced assessment.
Keywords: Learning Analytics, Educational Data Mining, Social Learning Environments,
Learning Management Systems, Technology-Enhanced Assessment TEA, TechnologyEnhanced Learning TEL, Informal Learning, Social Network Analysis.
Categories: H.3.1, H.3.2, H.3.3, H.3.7, H.5.1

Introduction

The use of technology in teaching and learning brings optimism and opportunity for
education, freeing both the teacher and the student in the facilitation of academic
knowledge, removing traditional boundaries and constraints, through the open and
ubiquitous access provided [Katz, 10]. Information Technology (IT) represents a set
of tools and applications that allow the incorporation and strengthening of new
educational strategies. The application of IT in the educational process is particularly
relevant. The interest of educators in using these technologies in the teaching
process presupposes greater engagement and an increase in student motivation in
understanding the content [Fonseca, 14], the educational innovation and the
incorporation of best practice in the educational process and test their effects on
learning and the assessment.
In the last few years, there has been a growing interest in the automatic analysis
of educational data to enhance the learning experience, a research area referred to
recently as learning analytics. The 2011, 2012, and 2013 Horizon Report identified
learning analytics as a possible key future trend in learning and teaching [Johnson,
13].
The learning environments have had an amazing evolution [Expertus, 10], from
the classic Learning Management Systems (LMS) to the environments of Educational
Social Networks, Social Learning Environments (SLE) and platforms MOOCs, with

features of cooperation, collaboration, and new types of learning, as the informal


learning.
The software tools provided by IT for the educational environments is in
evolution, there are lot of development projects at the level of architecture or
prototypes [Bichsel, 13], [Current, 13], primarily should be integrated into the
learning environments, features such as; learning analytics, techniques of educational
data mining, social network analysis to all forms of educational activities [Siemens,
13], with particular emphasis on the analytical of data educative to know
performance, learning and educational relations of students in their academic training,
to focus the extension of traditional methods of assessment to one based in the context
of technology-enhanced assessment. Therefore in this article we want to argue:
a) The formal and technical aspects of practices and techniques of Learning
Analytics in the learning environments. These practices and techniques are based on
the Social Network Analysis and the Educational Datamining. The learning
environments are referred to educational social networking and social learning
environments, but also could be done in any traditional LMS.
b) The proposal from an alternate assessment or complementary to the
traditional educational assessment, based fundamentally on mining of relationships of
the informal learning that establish the students in their research and studies activities
on a platform or social learning environment.

2
2.1

Literature Review
Learning Analytics and Educational Data Mining

Long and Siemens at the 1st International Conference on Learning Analytics and
Knowledge [Long, 11] defined that, learning analytics is the measurement, collection,
analysis and reporting of data about learners and their contexts, for purposes of
understanding and optimising learning and the environments in which it occurs. For
JISC-CETIS there is a meaning more generally [Powell, 12], analytics is the process
of developing actionable insights through problem definition and the application of
statistical models and analysis against existing and/or simulated future data.
NMC Horizon Report 2013 [Johnson, 13] has a more binding respecter with
education, the learning analytics is an emergent field of research in the education that
aspires to use data analysis as support-decisions on every area of the education, from
understanding student data to build better pedagogies, target at-risk students, and to
assess whether programs designed have been effective and should be sustained.
The Learning Analytics seen as process of connect-making of learning between
students: According to connectivism [Siemens, 04], the learning is the creation and
removal of connections between the entities, or the adjustment of the strengths of
those connections. So same for [Siemens, 12a], the importance of learning analytics is
on the point of view of learning, which is seen as a process of connect-making.
From a technical perspective [Chatti, 12], the support for Learning Analytics,
includes methods, such as statistical and visualization tools or Social Network
Analysis (SNA) techniques, and puts them into practice for studying their actual
effectiveness on the improvement of teaching and learning. Also, it actually borrows
from different related fields and synthesizes several existing techniques, as: academic

analytics, action research, educational data mining, recommender systems, and


personalized adaptive learning.
The Educational data mining has emerged as an independent research area in
recent years, starting with the first workshop on Educational Data Mining (EDM) was
held in 2005 in Pittsburgh, the 1st International Conference on Educational Data
Mining held in 2008 in Montreal, with the appearance of the Journal of Educational
Data Mining (JEDM) in 2009, the first Handbook of Educational Data Mining was
published in 2010, and finally in the summer of 2011, the International Educational
Data Mining Society (IEDMS) was formed to promote scientific research in the
interdisciplinary field of educational data mining , organizing the conferences and
journal, and the free open-access publication of conference and journal articles
[Chatti, 12], [Romero, 10], [Siemens, 12b].
According to [Romero, 10a], EDM is concerned with developing methods to
explore the unique types of data that come from an educational context and, using
these methods, to better understand students and the settings in which they learn.
From a technical perspective, EDM is the application of data mining techniques to
educational data (i.e. clustering, classification, and association rule mining) to support
teachers and students in analyzing the learning process, and so, its objective is to
analyze this type of data in order to resolve educational research issues and
understand the setting in which students learn [Chatti, 12].
Thus, we find that around of data and analytics in education, teaching, and
learning raises the priority for increased, high-quality research into the models,
methods, technologies, and impact of analytics, have emerged two research
communities - Educational Data Mining and Learning Analytics -. Therefore, it is
important for the researchers and educators to recognize the unique attributes of each
community, while that they share many attributes and have similar goals and interests,
they also have distinct technological, ideological, and methodological orientations
[Siemens, 12b].
Siemens and Baker [Siemens, 12b], in terms of the goal and the techniques and
models, locate some key differences between Learning Analytics and Educatinal Data
Mining as: The end goal of the learning analytics is to facilitate human judgement,
automated models play a part in helping humans make judgements, whereas that, for
the educational data mining the end goal is automated models, humans play a part in
setting these up. Similarly, the techniques and models in learning analytics, are the
social network analysis, sentiment analysis, influence analytics, discourse analysis,
learner success prediction, concept analysis, sensemaking models. Whereas the
techniques and models that are used for educational data mining, are the
classification, clustering, Bayesian modelling, relationship mining, discovery models,
visualization.
2.2

Technology-Enhanced Learning and Technology-Enhanced Assessment

The term Technology-Enhanced Learning (TEL) is used in the educational world to


describe the application of Information and Communication Technologies to learning
and teaching. The word enhancement suggests an improvement or betterment some
way, however, it is rare to find explicit statements about its meaning. A fundamental
question is, whether there is a generally accepted view of what constitutes learning in
higher education and how it can be enhanced? [Kirkwood, 14]. According to

[Manouselis, 11], the term TEL aims to design, develop and test sociotechnical
innovations that will support and enhance learning practices of both individuals and
organisations. It is therefore an application domain that generally covers technologies
that support all forms of teaching and learning activities.
Technology-Enhanced Assessment (TEA) it refers to the wide range of ways in
which technology can be used to support the assessment and feedback. TEA is the use
of the technology to improve the assessment, understanding this as the heart of the
learning experience, how a learner are assessed shapes their understanding of the
curriculum and determines their ability to progress [JISC, 10].
TEA includes on-screen assessment, often called e-assessment [JISC, 10].
According to [Wilhelm, 12] e-assessment describes the utilization of technology for
organization and execution of assessment activities in general, depending on the
degree of support that provide the environments. The relations e-Assessment and TEL
are preceded by the time. So same for [Timmis, 12], e-assessment also known as TEL
in one form or another, has been around for more than two decades.
In Although for [Timmis, 12] and [Wilhelm, 12] e-assessment and TEA is the
same, however, e-assessment is associated to the traditional management of the
evaluation in general, and TEA is a response to the emergence of educational
principles and practices associated at the types of assessment and yours feedback
respectively, and at the new technology-based tools, as tools of social media, social
web, semantic web and its application to education in their proposals for evaluation.

Method

This section presents the first point of our proposal, a framework based on the
techniques of the Social Network Analysis and Educational Data mining. We end the
section presenting the interfaces of a software prototype that implements this.
3.1

Fundamentals of a framework of learning environment with learning


analytics

Internet is a computer network and a set of services (V, E) with nodes V (estations,
hosts, servers) sending messages through of the links E (channels), E VxV. A
connection is a transmission link unidirectional or bidirectional, from a sender node to
a group of receiving nodes. The processes of concurrency, awarenness, access control
and resource management are duly substantiated by the Groupware and ComputerSupported Collaborative Work [Dommel, 00], [Hong-Tran, 03], [Sun, 02].
Definition 1.- A collaborative environment in a network G (V, E) is a tuple =
<S, U, R, F >. Where; S is a set of sessions, U is a set of users (hosts, processes,
agents, participants), R a set of shared resources, and F a set of protocols -floors
control- who control the resources.
Definition 2.- A collaborative session is a set of participamts Web sites
W1 ,..., W n
connected by the network on the environment of collaboration , with the condition
that there must be a user for each Web site. Any site can and should communicate
with sitio other site.

Accordin to [Dommel, 00], each site hosts an application which collaborates with
other applications of remote sites. Each application implements a set O of operations:
Op1 , Op 2 ,..., Op m . A Web site sends events to other sites by an operation
independently. When a site receives an event, identifies and executes the operation (or
operations) specified by the event [Sun, 96], [Chen, 09].
A typical example of collaborative application are the social networking, the
learning environments. If si y s j symbolize students at the sites W i y W j which
they are interacting solving their tasks. These students in their relations form a Social
Network. A social network of real life are complex systems [Butts, 01] that can be
represented traditionally by a formal model of graph and a data structure.
For the focus of our work we extend the model of simple graph to graph multinode, multi-mode, taking advantage of the contributions of Tang [Tang, 10] and San
Martin [San-Martin, 12] in their definitions of social networks extrapolated to social
learning environments. According to Jane Hart [Hart, 09] establishes that the heart of
a social learning environment is the social network. Thus, our formal proposal of
framework for social learning environments is based on well-founded work related to
learning environments and social networks [Tang, 10] , [Butts, 08], [Singh, 99], [SanMartin, 09] , [San-Martin, 12], [Gjoka, 11].
Definition 3.- A Social Learning Environment, is a multi-modal multi-relational
social network [Tang, 10], [San-Martin, 12] with attributes on actors and relation i.e.
a system
( N , E, LN , LE , , v, )
Where, the network is tripartite regarding the partition of the node set N in actors,
relations, and attributes:
1) The set of nodes N A T C such that A T C , and is a unin
of the actor set A , with the relation set T , and the attribute set C.
For the set of actors A , there exists a finite collection of families (subsets) of
actors A = A1 , A2 ,..., Ak such that every Ai A and ik1 Ai A .
For the set of relations T , there exists a finite collection of families (subsets)
of relations T = T1 , T2 ,..., T j such that every Ti T and ij1Ti T .
Both sets of actors A and relations T may be described using attributes.
2) The set E is a set of arcs (corresponding social ties, links) that admits a
partition in families of relations E EAT EAC ETC with EAT EAC ETC ,
where: E AT is the set of arcs between actors and relations, ETC is the set of arcs
between relations and attributes y E AC the set of arcs between actors and attributes.
3) LN is a set of node labels, where LN LA LT LC is a disjoint union of the set
of actor labels LA , with the relations labels set LT , and attribute labels set LC .
4) LE is a set of arc labels, where LE LAT LAC LTC is a disjoint union of the
set of labels of arcs between actors and relations LAT , with the set of labels of arcs
between actors and attributes LAC , and the set of labels of arcs between relations and
attributes LTC .

5) AT , AC , TC is a set of incidence functions, v vA , vT , vC is a set


of node labeling functions y AT , AC , TC is a set of edge labeling functions.
All arcs between the same pair (u, v) with u A and v T must have
different labels.
6) The following condition holds for all edges between the same pair of actors and
relations, for all e1 y e2 such that (e1 ) (e2 ) (u, v) with u A and v T ,

e1 , e2 E (e1 ) (e2 ) .
7) For a relation r T between two actors a1 , a2 A , such that there exist e1 , e2 E ,
and (e1 ) (a1, r ) , (e2 ) (a2 , r ) with labels (e1 ) p1 , (e2 ) p2 . The direction of r can
be specified by the ordered pair of participation labels, that is a direction
( p1 , p2 ) indicates that r starts in a1 and ends in a2 , the opposite direction is
represented by ( p2 , p1 ) .

Table 1: N studentscourses multi-nodes of social learning environment.

The Table 1 above shows a configuration of the data in the learning environment

( N , E, LN , LE , , v, ) , discloses the multi-modal multi-relational social


network, with multi-relations {friend-of, shared-with, enrrolled-in} and multiattributes. The actors: {Marquina, Benavides, Paima, Flores, Sotelo, Centurin,
Burgos, Paz}, are enrolled-in courses: {Artificial-Intelligence, Seminar-Thesis,
Complex-Systems, Data-Mining}. And the relationships friend-of describes the
degree of friendship that these students have, with the goal of share and collaborate
shared-with their study notes, class notes, and working papers of practices of
courses who share tuition.
Relation friend-of: In a social network real the labels on its links can be
numbers, for example, in the social network of students of adjacency matrix of the
table 1, if the label of the link is friend-of=1 between the nodes students
Centurion and Paz , could represent the attribute closest friend. So it friendof=2 between the students Centurion and Sotelo may represent the attribute
friend less friend or if friend-of=3 represent attribute distant friend. A social
learning environment thus represented, is a multi-mode social network [Tang, 10],
[San-Martin, 12].
Relation enrolled-in: Similarly the same social learning environment may be
a multi-node social network [Tang, 10], [San-Martin, 12], i.e. the links (of the
relation) may be established between nodes of different types, for example in the
figure 1, the corresponding nodes to students Centurion, Paz and Sotelo are
related through the link enrolled-in to node course="Artificial Intelligence, as well
as the node Sotelo also has the link enrolled-in with the Seminar node. The
figure 1 is an example of multi-node social network on the relationship enrolled-in
between student nodes and subject (course) nodes.

Figure 1: Social learning environment is a multi-node, multi-node social network


with relations friend-of, and "enrolled-in. Adapted from [San-Martin, 12]
Graph database. Neo4j is used as a graphical database [Neo4j, 14] to store
information of the relations and graphs of students on enrollment in their courses or
sharing their learning materials them accurately and find it later. The data model is the

graph. We use Cypher as graph query language, which is a declarative query


language [Cypher, 14]. The graph of the figure 2 Part (a), shows at the students
s1 , s2 , s3 , s4 , s5 , s6 s7 , s8 enrolled in the subjects (link enrolled-in) a1 =ArtificialIntelligence,

a2 , a3 , a4 . The Part (b) shows the creation of the graphical database

DataBase_description.txt. The node AI is created and labeled as Course with


their corresponding attributes AI : Course name : ' ArtificialIntelligence ',credit : 3 as
well as their related nodes. An example of student enrolled in a subject, take the
students case s1 =Marquina enrolled-in in the subject AI, where the format in
Neo4j is: ( gustavoM ) [: ENROLLED IN ] ( AI ) .

Figure 2: Part (a) Social network generated by relationship enrolled-in. Part (b)
Portion of the graphical data base DataBase_description.txt in Neo4j.

Tracing Graph. Any of the enrolled-in, friend-of, shared-in relationships


can be mathematically represented by a graph. Fruchterman and Reingold
[Fruchterman, 91] present an algorithm for tracing graph -algorithm FR(Graph G )based on two esthetic criteria: a) The nodes generally should not be drawn too near
each other, and b) The nodes connected by an arc they should be drawn near.
The algorithm is based on molecular or planetary simulations; nodes behave as
atomic particles or celestial bodies, exerting attractive and repulsive forces on the
others, similar to Hooke's law. The time complexity of the algorithm is O(n3 ) .
Let G ( N , E ) be, an undirected graph that represent to any of the relationships
enrolled-in, friend-of and shared-in, with N the set of vertex ( N n) and E
the set of edges (u, v) / u, v N , u v , with E m . pv ( xv , yv ) is the vertex

pu pv y
duv is the distance between
pu pv v

position v . So, pu pv is the unit vector


the positions of the vertex

u and v , with duv pu pv pu pv 2 .


2

Definition 4.- The force f (u ) that undergoes a vertex u ( with respect to vertex
v and (u, v) / u, v N , u v ) for tracing graph G , the algorithm FR(Graph G ) is:

f (u )

( u ,v )E

f a (u, v)

vV ,u v

f r (u, v)

where:
2

a) The attractive force is: f a (u , v )

d uv
K

pu pv , with K C

an experimental

constant that represents the ideal distance between the vertices of the graph so that
they are uniformly distributed, C is an experimental constant, n N and a is the
area available to draw the graph.
2
b) The repulsive force is: f (u, v) K p p , for all
r
u
v

u, v V ( u v ).

duv

Graph Visualization. The visualization of the graphical database [Fionda, 14] of


interaction conducted in the social learning environment, of the relationships
enrolled-in, friend-of, shared-in between the actors {Marquina, Benavides,
Paima, Flores, Sotelo, Centurin, Burgos, Paz, } respect to be enrolled in the
subjects {Artificial-Intelligence, Seminar-Thesis, Complex-Systems, Data-Mining},
be friends, or share the educational materials {working papers of practices, study
notes, class notes} respectively, is performed by the following process illustrated
in the figure 3.
Step1. The host1 GRAPHENEDB [GrapheneDB, 14] (add-on of HEROKU [heroku,
14]) has stored the graphical database NEO4J DataBase_description.txt of the
students {Marquina, Benavides, Paima, Flores, Sotelo, Centurin, Burgos, Paz, }

1
GRAPHENEDB is an add-on" of Heroku and is the host of the databases Neo4j. The prototype of
social learning environment SGroupMeeting Web APP SGroupMeeting is being developed with the
framework web RAILS of the Ruby language, and is housed in the platform HEROKU as a service.

enrolled in the subjects { Artificial-Intelligence, Seminar-Thesis, Complex-Systems,


Data-Mining } of the figure 2, Part (b).
Step2.
Neography
[Marzi,
14]
retrieves
the
database
Neo4j
DataBase_description.txt by means of REST API and returns it as JSON
response.
Step3. The retrieved information is processed by the library VivaGraphJS
[Kashcha, 14] - incorporating the algorithm of tracing graph FR(Graph G )
[Fruchterman, 91] - for visualization as graph. [Neo4j, 14].
Analytical based on Social Network Analysis (SNA). The structural measures
and indices of centrality of SNA allow us to understand and measure the performance
of students in their learning relationships [Borgatti, 05].
Definicin 5.- Any of the relationships E =shared-in (friend-of and sharedin) that establish the students s N ={Marquina, Benavides,} each other in a

Figure 3: Process of retrieval & deployment of databases graphic Neo4j to visualize


information from the social networks formed by the relations enrolled-in, etc.
collaboration session on social learning environment ( N , E, LN , LE , , v, )
can be described by a simple graph G ( N , E ) and is subject to the following:
a) A path from s N to t N is an alternate succession of nodes and links,
starting at s N and terminating at t N , such that each link connects your
precedent with their subsequent.

b) The minimum length of any path connecting s y t in G , denoted by


dG ( s ,t ) , defines the distance between the vertices corresponding to the students s
and t N , with dG ( s,t )dG (t ,s ) and dG ( s,s )0 .
c) Let st be the number of shortest paths from s N to t N , such that
st ts and ss 1 . Let st (v ) denote the number of shortest paths from s to t for
any node v N that lies in the path between s and t . For example, in the relations of
the students in , the measure st (v ) for s =Paima, t =Sotelo and v =Benavides,
shows us the optimal communication paths of the students Paima and Sotelo in which
it participates Benavides.
d) The centrality indices of a node v N , they are:
1
1
(Closeness), CG (v)
(Degree Centrality),
C (v )
maxtV dG (v,t )
tV dG (v,t )
C
st (v)

st (v) (Stress Centrality), C (v )


(Betweenness).
C (v )
S
B
s vtV st
s vtV

The centrality indices of v N are interpreted as the role that students have in the
social structure, being therefore a structural attribute, i.e. a value assigned that
depends strictly on its location in the network [Borgatti, 05]. For implementation of
centrality indices - BetweennessCentrality(vN ) - is chosen algorithms with temporal and
spatial complexity O( nm) and O( nmn2 log n) for n y m number of nodes and
number of links [Brandes, 01].
Recommendation and Reputation [Christensen, 2014]. The recommendation of
the documents or materials informal learning working papers of practices, study
notes and class notes, that students share through social learning environment, it
can help customize the learning scheme for each student, using the preferences or the
automatic recommendation of these documents.
The set of nodes N A T C of the social learning environment ,

A s1 , s2 ,..., sn is the set of actors students that interact on-line, and the relation

Ti shared-with as illustrated in the figure 4(a), of the set of relationships T may


be described using attributes labels, 1: working papers of practices, 2:study notes
and 3:class notes that correspond to the arcs between relations and attributes, so
ETC {working papers of practices, study notes, class notes} of
E EAT EAC ETC in

Therefore, if shared-with=1 between the nodes students Centurion and Paz


or Paima and Sotelo, represent that they are sharing working papers of
practices. Similarly, if shared-with=2, between the students may represent that
they are sharing study notes and if shared-with=3 are sharing class notes of the
various subjects that are enrolled.
A d document can be any of the types {working papers of practices, study
notes, class notes} and can be in any format PDF, JPG, PNG, RDF, DOC. A d

document is described by a vector of keywords d (k1 ,k2 ,kn ) , for example in


the figure 4(b) the document that share the students, have the vector of keyword d =
(page-rank, frobenius, search-web, mining-web). In 4(c) it can be seen the social
network of the relationship shared-with=1 of the figure 4(a) sharing the documents
of the ype working papers of practices.

Figure 4: In (c) the social network of students sharing document working papers of
practices of keywords d = (page-rank, frobenius, search-web, mining-web) de (b).
The record of behavior of the actors si A respect to documents d are vectors of
behavior x (si , Behavior, BehaviorTime,d ) , where Behavior {browsing, upload,
download} are the basic actions on-line, that perform the students si A respect to
documents d , and BehaviorTime is the time elapsed during the execution of Behavior .
The search for a document d j in the set of all documents D d1 , d2 ,..., d N that
are shared on social learning network by keywords (k1j ,k2j ,kmj ) using text-

mining [Aggarwal, 12], it means reduce the impact of the intervention of simple terms
that are not keywords, but they have been entered into the search (for example: the,
of, on, etc.). To this, the weight wij of each keyword ki j of the document d j is
calculated and is indexed to vector of weights ( w1j ,w2j ,wmj ) .
Definition 5.- The weight w of each keyword k of a document d , is an
application w(k , d ) and is calculated by TF-IDF (Term Frecuency Inverse
Document Frecuency) [Chung, 08], [Paik, 10]
tf (k , d ) log(
w(k , d )

N
0.01)
nk

(tf (k , d ) log( n
k 1

0.01)) 2

Where, tf (k , d ) is the frequency of the kth keyword k in the document d , N


is the total number of documents, nk is the frequency of the kth keyword k of the
document d that appears in the set of all documents D .
Thus, the ranking and recommendation of the documents d1 , d2 ,..., d N accessed
by search of keywords using text-mining, can be done using similarity metrics of
documents Sim(di , d j ) , based on the cosine of the inner product of d i and d j on
their vectors of indexed weights [Adomavicius, 05].
Definition 6.- For any pair of documents di , d j D described for their vectors
of keywords (k1i ,k2i ,kmi ) and (k1j ,k2j ,kmj ) and their indexed vectors of
weights ( w1i ,w2i ,wmi ) and ( w1j ,w2j ,wmj ) respectively. The similarity of
d i and d j is obtained by:

Sim(di , d j )

k 1, m

k 1, m
i 2
k

(w )

wki wkj

k 1, m

( wkj ) 2

So, we can draw up a list of recommendation of documents (di );i 1, N by


similarity and the rankings rank ( si ) of its authors Si ;i 1, n and implement them by
the algorithm Recommendation-Engine( di D , si A ) [Leony, 13].
3.2

Application of the Framework

As application of framework of learning environment with learning analytics is


being developed the artifact SGroupMeeting, acronym of Social Group Meeting, a
prototype software that implement part of the features outlined in the Section 3.1 for
testing of the present work. SGroupMeeting this at an early stage of development,
looking to implement: a) the educational social network [Bricen, 13] for the purposes
of to support the informal learning, and share documents of educational activities
(papers, class notes, etc.), b) functionalities of learning analytics with the engine
Analytical Educative, which presents a viewfinder of indicators of centrality of the
social network of students, as well as the assessment, c) functionalities of the
educative mining as reputation, clustering, association and recommendation, d)
Search graph and graphic database, tracing and visualization.

Results

In this section we present the proposal of assessment based on social network analysis
that establish students in the development of their academic activities on-line
4.1

Tools, Users and Assessment

Analytical Educative is a tool viewfinder of the assessment - a functionality of the


prototype of learning environment SgroupMeeting that is under development implements some of the techniques of social network analysis and educational data
mining, as learning analytics capabilities, level indicators and metrics for student
assessment [Coates, 13]. It gets posing the cursor over each student node of the social
network in analysis.
Analytical Educative, presents a summary of: a) Data of student node attributes of node in the current network-; b) Information of the linked data of your
nodes in the social network; c) Data of e-Activity: working papers, study notes, notes
of course -associated with the current subject (or course) and the ralations of
collaborations-; d) Data of mining: reputation, recommendation; e) Data of social
network analysis: Centrality - betweenness, closeness, degree - and cohesiveness-; f)
The ASSESSMENT of the student or centrality-based assessment.
So, the viewfinder presents the data analytic of the students. In the figure 5 shows
an example of the data analytical of the student Paima G. and its assessment in the
development of the practical#2AI.

Figure 5: Snapshot of viewfinder of the data analytic of the student Paima.


In a typical scenario - of collaborative relationships of a social network, as the of
figure 4 (shared-in of attribute 1:working-papers) of the group of students G=

{Marquina, Benavides, Paima, Flores, Sotelo, Centurin, Burgos, Paz} wich share the
working paper of keywords d =(page-rank, frobenius, search-web, mining-web)
solution of the practice#2AI -, the assesment, is calculated as:
ASSESSMENT ( si )

10 NormCloseness 60 NormBetweenness 10 NormGrade

100

rank ( si )

Where: rank ( si ) is the normalized values of ranking of students Si ;i 1, n author of


documents d j , NormGrade, NormCloseness y NormBetweenness, are the normalized
values of Grade, Closeness y Betweenness of the students Si ;i 1, n , in percent and
scale [0.0-5.0] - the scale is explained in the next section -. It is seen, that has been
weighted to the leadership provided by the indicator Betweenness as a bridge of
connectivity between students and between clusters or communities of learners, and
has been discriminated the indicator on Degree to the being an egocentric local
measure. However the criteria for assigning weights of the indicators of centrality,
must be a configurable option for the user for the purposes of assign weights
according to the indicator you want to highlight (this is a feature that is still
developing)

Discussion

We seek to verify the hypothesis [Groebner, 11]: the assessment based on learning
analytics is as reliable and robust than that done by traditional methods in the context
of this work. This expressed in terms of statistical hypothesis could be stated as
follows H: that the proposed evaluation system with learning analytics is equal to or
better than the current traditional assessment system in use.
For the verification we developed a test design with the current group G of n
students, observing and evaluating the results of the development of practice#1AI and
practice#2AI in the course of Artificial Intelligence. We configure the data collection
into two stages: One Pre-Test mode 1: That involves taking the data of evaluation at
the Group G using the traditional or conventional system of collaborative group
learning in the development of practice#1AI; and the other Post-Test mode 2: That
consist of evaluating the G Group applying the proposed with learning analytics new
system (ie with the analysis of the relations of cooperation and collaboration, over a,
on-line learning environment) in the development from practice#2AI. The objective is
to evaluate the variable learning analytics in an environment of on-line learning and
its relation with the assessment.
The evaluation indicators of academic performance that were used are those of the
relationship Value-range:Sufficiency-Indicator ([0-0.75]:Disapproved <0.75-1.5]:
Regular, <1.5-2.25]:Approved, <2.25-3.25]:Outstanding, <3.25-5.0]:Excellence). The
average values of evaluation of Pre-Test and Post-Test raised by the subject teacher of
Artificial Intelligence in the practice#1 and practice#2 respectively, are described on
the left side of table 2.
Likewise, for the hypothesis H, the statistical hypotheses referred [Groebner, 11],
[Weaver, 11] are; the null hypothesis H0: The current (or traditional) evaluation
system is better than the proposed system with learning analytics (H0: 120), and

the alternative hypothesis H1: The proposed system of assessment with learning
analytics is better than the traditional system (H1: 21> 0).
Group GStudent
Marquina
Benavides
Paima
Flores
Sotelo
Centurion
Burgos
Paz
D. Estandar

Post-Test

Pre-Test

Practice#2AI

Practice#1AI

2,7
2,2
4,5
3
4,3
1,8
3,2
4,5

1,5
0,4
3,0
1
2,8
0,5
1,9
3,1

Dif.
D
1,2
1,8
1,5
2,0
1,5
1,3
1,3
1,4

0,273

Table 2: Comparison of means for dependent observations, with Student's t-Dist.


To verify whether the H hypothesis is accepted or rejected [Weaver, 11],
[Loveland, 13], using data from the table above, we use the Student's t-Distribution,
which is a probability distribution that arises of the problem of estimating the average
of one normally distributed population when the size of the sample n is small. The
parameter n-1 is the number of degrees of freedom.
As the conclusions we arrived are based on a sample n, it is possible be mistaken
and take the incorrect decision reject the null hypothesis H0 when it is true,
statistically is said to be an error of type 1. The probability of making a Error of type
1, known as significance level and is the size of the rejection region, the
complement of the rejection region is called the level of confidence. In distribution
Students of the table 2, the graphic to the right we illustrates the data applicable to
this case. The rejection region is the set of values such that if the statistical test test
statistic- falls within this range, we decided to reject the null hypothesis H0.
For the data of the table 2, with n = 8, and standar deviation D 0,273, taking a

significance level of 0,05 (size of the rejection region) and degrees of freedom
(n-1) = 7, the critical value in the table t-student is tt 1,895, as shown in the
graphic to the right of table 2.
D
The calculation of the test statistic t-student, is te
= 15, 46. Where it
D / n
is observed that te = 15, 46 > 1,895= tt , i.e. the test statistic has fallen within the
rejection region, shown in the graphic of the Student's t-Distribution of the table 2 and
therefore we conclude that there is sufficient statistical evidence to infer that the null
hypothesis H0 is false and thus take the alternative hypothesis H1: The proposed
system of assessment with learning analytics is better than the traditional system as
true.
Consequently, we accept the hypothesis the assessment based on learning
analytics is as reliable and robust than that done by traditional methods posed at the
beginning of this section.

Conclusions

This work suggests that it is possible the individualization and customization of the
functions or services educationals for each student. Not only in the support for the
delivery of content, if not also into something that is very sensitive: the assessment.
So that methods and technologies of the framework of learning analytics of this paper,
sustain that the assessment of each student responding to their performance and
individual characteristics, it is possible, without disregarding the relationship with
their peers from group or his companions current educational network.
Also, the restriction of this worker this in that the assessment results are preliminary,
based only on indicators of centrality, cohesion and reputation of the students in the
current social network. Lack unify, academic and pedagogical criteria, of the
traditional assessment formativa, sumativa-, datos de mineria, types de assessment
massive peer, self and eAssessment- and centrality-based assessment and combining
all these data, in formulas of measures of dispersion, adjustment data and
normalization to pass toward indicators of assessment, that signifying the promotion
or disapproval of the subject or academic course, from a more general view.
The data shown in Section 5 for verification and testing are referred to a single test
group, the data from the other 7 groups that were sampled, also threw similars results,
hence it is still working on the implementation, monitoring and analysis of other
students groups, diversifying the relations of the social network and the types of
assessment, whose results will be published in a companion article. However, these
tests provide an important fact, that using Student's t-Distribution, allows us to infer
population mean results from the analysis of sample mean. Consequently, it can be
established that the assessment with learning analytics is more reliable than the
traditional evaluation, for the entire student population of the subject or course within
a learning environemnt. Therefore it can be concluded that learning analytics is a
reliable element and may be included in the practices of technology-enhanced
assesment.

Future Work

Continue investigating on the same line, in educational mining and learning analytics
for personalized learning environments and connected through the linked-data
education.

References
[Adomavicius, 05] Adomavicius, G., Tuzhilin, A.: Toward the Next Generation of
Recommender, Systems: A Survey of the State-of-the-Art and Possible Extensions, IEEE
Transactions on Knowledge and Data Engineering, vol. 17, no. 6, June 2005.
[Aggarwal, 12] Aggarwal, Ch., Zhai, Ch.: Mining Text Data, Springer, 2012.
[Bichsel, 13] Bichsel, J.: The State of E-Learning in Higher Education: An Eye Toward Growth
and Increased Access (Research Report), Louisville, CO: EDUCAUSE Center for Analysis and
Research, June 2013, available from http:/www.educause.edu/ecar

[Borgatti, 05] Borgatti, S.: Centrality and network flow, Social Networks 27, pp. 55-71, 2005.
[Brandes, 01] Brandes, U.: A Faster Algorithm for Betweenness Centrality, Journal of
Mathematical Sociology vol. 25, issue 2, pp.163-177, 2001.
[Bricen, 13] Bricen, H., Uzunboylu, H.: The Use of Social Networking Sites in Education, A
Case Study of Facebook, Journal of Universal Computer Science, vol.19,5, pp.658-671, 2013.
[Butts, 01] Butts, C.: The complexity of social networks: theoretical and empirical findings.
Social Networks, 23:31-71, 2001.
[Butts, 08] Butts, C.: Social Network Analysis: A Methodological Introduction, Asian Journal
of Social Psychology, vol. 11, 13-41, 2008. DOI: 10.1111/j.1467-839X.2007.00241.x
[Chatti, 12] Chatti, M., Dyckhoff, A., Schroeder, U., Ths, H.: A reference model for learning
analytics, International Journal of Technology Enhanced Learning, vol. 4, no 5, 2012.
[Chen, 09] Chen, D., Sun, Ch.: Context-Based Operational Transformation in Distributed
Collaborative Editing Systems, IEEE Transactions on Parallel and Distributed Systems,
Volume 20, pp. 1454-1470, 2009.
[Christensen, 2014] Christensen, I., Schiaffino, S.: A Hybrid Approach for Group Profiling in
Recommender Systems, Journal of Universal Computer Science, vol. 20, no. 4, 2014.
[Chung, 08] Chung, H., Pong, R., Fai, K., Lam, K.: Interpreting TF-IDF Term Weights as
Making Relevance Decisions, ACM Transactions on Information Systems, Vol. 26, No. 3,
Article 13, 2008.
[Coates, 13] Coates, H., Wilkinson, D., Edwards, D., Canny, B., Pearce, J.: "Developing
outcomes assessments for collaborative, cross-institutional benchmarking: Progress of the
Australian Medical Assessment Collaboration", Medical Teacher, 2013.
[Current, 13] Current State of Learning Analytics in Higher Education. Drivers, Trends,
Benefits, and Challenges. Learning Analytics in Higher Education: An Annotated
Bibliography, EDUCAUSE 2013, available from: http://www.educause.edu/visuals/shared/
events/ELI149/learning-analytics-in-higher-education-annotated-bibliography.pdf
[Cypher, 14] Cypher: Neo4js Query Language, Neo Technology, Inc., 2014. Available from
http://neo4j.com/developer/cypher-query-language/
[Dommel, 00] Dommel, H., Garcia, J.: A Coordination Framework and Architecture for
Internet Groupware, Journal of Network and Computer Applications, Special Issue on "Support
for Open and Distance Learning on the WWW", Academic Press, Volume 23, Issue 4, pages
401-427, October 2000.
[Expertus, 10] Expertus and Training Industry, Inc.: Survey Results: The Current and Future
State of Learning Management Systems, 2010, available from: http://www.trainingindustry.
com/media/3314559/the%20state%20of%20lms%20report.pdf
[Fonseca, 14] Fonseca, D., Mart, N., Redondo, E., Navarro, I., Snchez, A.: Relationship
between student profile, tool use, participation, and academic performance with the use of
Augmented Reality technology for visualized architecture models, Computers in Human
Behavior, vol. 31, pp 434-445, February 2014.
[Fionda, 14] Fionda V., Gutierrez C., Pirr G.: Knowledge Maps of Web Graphs, in the
proceedings of the 14th International Conference on Principles of Knowledge Representation
and Reasoning (KR2014), Vienna, Austria, 20-24 July 2014.

[Fruchterman, 91] Fruchterman T., Reingold E.: Graph Drawing by Force-directed Placement,
Software - Practice and Experience, 21(11), pp. 1129-1164, 1991.
[Gjoka, 11] Gjoka, M., Butts, C., Kurant, M., Markopoulou, A.: Multigraph Sampling of
Online Social Networks, IEEE Journal on Selected Areas in Communications (special issue on
Measurement of Internet Topologies), forthcoming, 2011.
[GrapheneDB, 14] GrapheneDB: Neo4j Graph Database as a Service, heroku add-ons, 2014.
Available from https://addons.heroku.com/graphenedb
[Groebner, 11] Groebner, D., Shannon, P., Fry, P., Smith, K.: Business Statistics, 8/E
Introduction to Hypothesis Testing, Chapter 9, Pearson Education, New Jersey, 2011.
[Hart, 09] Hart, J.: Building a Social Learning Environment - For Free or at Low Cost, Inside
Learning Technologies, pp. 19-23, 2009.
[Heroku, 14] Heroku: Deploying applications, Heroku Dev Center, 2014. Available from
https://devcenter.heroku.com/articles/how-heroku-works#deploying-applications
[Hong-Tran, 03] Hong-Tran, M., Raikundalia, G., Yang, Y.: Methodologies and Mechanism
Design in Group Awareness Support for Internet-Based Real-Time Distributed Collaboration,
Web Technologies and Applications, Lecture Notes in Computer Science Volume 2642, 2003,
pp. 357-369, 15 Apr 2003.
[Johnson, 13] Johnson, L., Adams Becker, S., Cummins, M., Estrada, V., Freeman, A., and
Ludgate, H.: NMC Horizon Report: 2013 Higher Education Edition. Austin, Texas: The New
Media Consortium, 2013.
[JISC, 10] JISC- Joint Information Systems Committee: Effective Assessment in a Digital
Age: A guide to technology-enhanced assessment and feedback, JISC Assessment Symposium
2010, Higher Education Funding Council for England - HEFCE, 2010, available from:
http://www.jisc.ac.uk/media/documents/programmes/elearning/digiassass_eada.pdf
[Katz, 10] Katz, R. N. Scholar, scholarship, and the scholarly enterprise in the Digital Age.
EDUCAUSE Review, pp. 4456, 2010.
[Kashcha, 14]: Kashcha, A.: VivaGraphJS, Graph drawing library for JavaScript, GitHub,
2014. Available from https://github.com/anvaka
[Kirkwood, 14] Kirkwood, A., Price, and L.: Technology-enhanced learning and teaching in
higher education: what is enhanced and how do we know? A critical literature review.
Learning, Media and Technology, 39(1), 636, 2014.
[Leony, 13] Leony, D., Parada, H., Muoz, P., Pardo, A., Delgado, C.: A Generic Architecture
for Emotion-based Recommender Systems in Cloud Learning Environments, Journal of
Universal Computer Science, vol. 19, no. 14, pp. 2075-2092, 2013.
[Long, 11] Long, P., Siemens, G.: Conference LAK 2011 Alberta-Canada, Proceedings of the
1st International Conference on Learning Analytics and Knowledge, ACM, New York, 2011.
[Loveland, 13] Loveland, J.: Mathematical Justification of Introductory Hypothesis Tests and
Development of Reference Materials, Thesis MScs Mathematics, Utah State University, 2013.
[Manouselis, 11] Manouselis, N., Drachsler, H., Vuorikari, R., Hummel, H., Koper, R.:
Recommender Systems in Technology Enhanced Learning. In Recommender Systems
Handbook (pp. 387415), 2011. Springer. doi: 10.1007/978-0-387-85820-3
[Neo4j, 14] Neo4j: Graph Database Neo4j, Neo Technology, Inc., 2014. Available from
http://neo4j.com/developer/graph-database/

[Paik, 10] Paik, I., Mizugai, H.: Recommendation System Using Weighted TF-IDF and Naive
Bayes Classifiers on RSS Contents, Journal of Advanced Computational Intelligence and
Intelligent Informatics, Vol.14 No.6, 2010
[Powell, 12] Powell, S., MacNeill, S.: Institutional Readiness for Analytics, JISC Cetis - Centre
for educational technology & interoperability standards, Analytics Series vol.1, issue 8, The
University of Bolton, 2012. URI: http://publications.cetis.ac.uk/2012/527
[Romero, 10] Romero, C., Ventura, S., Pechenizky, M., Baker, R.: Handbook of Educational
Data Mining. 2010. Editorial Chapman and Hall/CRC Press, Taylor & Francis Group. Data
Mining and Knowledge Discovery Series, 2010.
[San-Martin, 09] San-Martin, M., Gutierrez, C.: Representing, querying and transforming social
networks with RDF/SPARQL, ESCW2009, 2009.
[San-Martin, 12] San-Martin, M.: A Model for Social Networks Data Management, Ph.D.
Thesis, Universidad de Chile, 2012.
[Siemens, 04] Siemens, G.: Connectivism, A Learning Theory for the Digital Age, 2004, online. http://www.elearnspace.org/Articles/connectivism.htm
[Siemens, 12a] Siemens, G.: Change MOOC: Sensemaking and Analytics, Technology
Enhanced Knowledge Research Institute (TERKRI), 2012.
[Siemens, 12b] Siemens, G., Baker, R.: Learning analytics and educational data mining:
towards communication and collaboration, Proceedings of the 2nd International Conference on
Learning Analytics and Knowledge - LAK 2012, Simon Buckingham Shum, Dragan Gasevic,
and Rebecca Ferguson (Eds.), ACM, New York, USA, pp. 252-254, 2012.
[Siemens, 13] Siemens, G., Shaun-Joazeiro, R.: State of the Field: Learning Analytics and
Educational Data Mining, EDUCAUSE, 2013, on-line.
[Singh, 99] Singh, L., Beard, M., Getoor, L., Blake, M.: Visual mining of multimodal social
networks at diferent abstraction levels. In Informatiion Visualization IEEE Computer Society,
pp. 672-679. IEEE Computer Society, 2007.
[Sun, 96] Sun, Ch., Yang, Y., Zhang, Y., David Chen, D.: Distributed Concurrency Control in
Real-time Cooperative Editing Systems, Proceedings Concurrency and Parallelism,
Programming, Networking, and Security: Second Asian Computing Science Conference,
ASIAN '96, Singapore, December 2-5, 1996.
[Sun, 02] Sun, Ch., Chen, D.: Consistency maintenance in real-time collaborative graphics
editing systems. ACM Trans. Computer-Human Interaction 9, 1 March 2002.
[Tang, 10] Tang, L., Liu, H..: Community Detection and Mining in Social Media, Morgan &
Claypool Publishers, 2010
[Timmis, 12] Timmis, S., Oldfield, A., Broadfoot, P., Sutherland, R.: Where is the cutting edge
of research in e-Assessment? Exploring the landscape and potential for wider transformation.
CAA 2012 International Conference, University of Southampton, UK, 2012.
[Weaver, 11] Weaver, B.: Hypothesis Testing Using z- and t-tests, Lakehead University,
Canada, 2011.
[Wilhelm, 12] Wilhelm, P.: A Generic Platform for Open Assessment Management in Higher
Education: A modular architecture for assessment management within a centrally hosted
learning and teaching portal, Diplom-Informatiker, Grades eines Doktors, Aachen University,
2012.

You might also like