A Personalized Learning Recommendation System Architecture for
Learning Management System
Thoufeeq Ahmed Syed
, Vasile Palade
, Rahat Iqbal
and Smitha Sunil Kumaran Nair
Middle East College, Muscat, Omam
Coventry University, Coventry, U.K.
Keywords: Technology Enhanced Learning, Learning Management System, Personal Learning Recommendation
Systems, Dataset, Educational Repositories.
Abstract: The information on the web is ever increasing and it is becoming difficult for students to find appropriate
information or relevant learning material to satisfy their needs. Technology Enhanced Learning (TEL) is an
area which covers all technologies that improve students learning. Effective Personal Learning
Recommendation Systems (PLRS) will not only reduce this burden of information overload by
recommending the relevant learning material to the students of their interest, but also provide them with
“right" information at the “right" time and in the “right" way. In this paper, we first present a detailed
analysis of existing TEL recommendation systems and identify the challenges that exist for developing and
evaluating the datasets. Then, we propose an architecture for developing a PLRS that aims to support
students via a Learning Management System (LMS) to find relevant material in order to enhance student
learning experience. Also we proposes a methodology for building our own collaborative dataset via
learning management systems (LMS) and educational repositories. This dataset will enhance student
learning by recommending learning materials from the former student’s competence qualifications. The
proposed dataset offer information on the usage of more than 19,296 resources from 628 courses apart from
data from social learner networks (forums, blogs, wikis and chats), which constitutes another 3,600 stored
files Finally, we also present some future challenges and a roadmap for developing TEL PLRSs.
Technology Enhanced learning is the application of
information and communication technologies to
teaching and learning (Kirkwood and Price 2014).
Recommendation Systems (RS) are software tools
based on machine learning and information retrieval
techniques (Aamir and Bhusry, 2015) that provide
suggestions for potential useful items to someone’s
interest (Ricci et al., 2011). RSs’ are widely used in
many fields including TEL (Verbert et al., 2012).
Until recent years Learning Management Systems
(LMS), a subset of TEL, had not been personalized.
Several researchers working in the field of LMS to
enhance students learning experiences highlighted
the need of RSs’ for LMS, so as to address the
following challenges in LMSs’:
Difficulty in sharing the learning resources;
High redundancy of learning material (Shishehchi
et al., 2011);
Personalization of information (Fischer, 2011);
Information overload (Manouselis et al., 2009)–
which is the ever increasing volume of digital
information particularly on the web, and due to
this reason it has becomes extremely more and
more difficult for learners to find suitable items to
satisfy a particular need;
Learning isolation.
Traditional LMSs only describe the basic
information learning resources but could not describe
the complex relationship between resources,
teachers, students and their peers (Poorni et al.,
2014). Also such systems could not integrate
learning from formal, informal and social network
learners (Fazeli et al., 2014).
This paper introduces a novel architecture for RS
via LMS, Moodle in our case. Learning-related
search problems are highly recurrent across a group
of students participating in similar teaching activities.
The process of discovering the optimal search query
Syed T., Palade V., Iqbal R. and Nair S.
A Personalized Learning Recommendation System Architecture for Learning Management System.
DOI: 10.5220/0006513202750282
In Proceedings of the 9th International Joint Conference on Knowledge Discovery, Knowledge Engineering and Knowledge Management (KDIR 2017), pages 275-282
ISBN: 978-989-758-271-4
2017 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
in a search engine like Google is repeated by all
students participating in a particular module and the
knowledge of the most optimal search query is not
captured or shared across the community of students
even though sharing this knowledge would
significantly increase the learners’ performance
(Zaina et al., 2010). Recommendation systems in
education filed should be personalized by the
objectives of the task rather than on the preferences
of students (Losada and Martín, 2014). Online social
networks allow users to share ideas, activities,
events, and interests within their contact network
(Prates et al., 2013). This research project will prove
the feasibility of using recommendation systems and
social media applications in educational
environments to deliver a working prototype of a
recommender system that will assist the learner, and
this implies learning from or reusing previous
students querying experience, activity-based context
which is based on users actions (Kramár and
Bieliková, 2012), use of social media interactions
among students. In section 2 we presents state-of-the
art and current RSs’ architectures and introduces
personalized learning RS for TEL. In Section 3 we
present the proposed PRS architecture. Section 4
presents the Datasets for TEL RSs’ and associated
challenges. Section 5 presents the challenges in
evaluating the RSs’ for TEL. Finally section 6
presents conclusions and direction for future work.
2.1 Recommendation Systems
There has been massive growth in research on RSs’
and the applications of RSs to various areas in the
last 15 years (Candillier et al., 2009). The following
are the commonly proposed approaches for building
2.1.1 Content based Filtering
Content based RSs’ predict an item to a user based
on the similarity between the items content
description and the user’s preferences model
(Pazzani and Billsus, 2007). Content based RSs’
work with user and item profiles which were referred
in the past. The referred profiles are represented as
vectors, which hold the characterizing attributes of
items or users. (del Losada and Martín, 2014) has
developed a Prototype of Content-Based RS in an
Educational Social Network. The experiments were
carried on in a real time environment ClipIt (elgg,
2016) tool. However, the recommendations
generated are uniform for all the students and lacks
. To enhance students learning experience
(Hsu, 2008) has developed an online personalized
English learning RS which facilitates English as a
Second Language (ESL) student with reading lessons
that suit individual student preferences.
2.1.2 Collaborative Filtering
Collaborative filtering techniques are most
commonly used in social networking and social
media environments for proposing user interactions
or interesting shared resources (Karampiperis et al.,
2014). Collaborative RSs’ aims to predicts individual
preferences and provide suggestions for links for
further resources or other systems, products and
resources which are likely to be of interest. (Dascalu
et al., 2015) developed an educational collaborative
filtering recommender agent (U Learn) with a build-
in integrated learning style recommender. Such RS
helps in providing suggestions and shortcuts for
learning materials and learning tools, helping the
learner to better navigate through educational
resources. The limitation of such RS are
recommendations made by the learners with similar
learning styles.
2.1.3 Knowledge based Filtering
Knowledge-based RSs integrate the user knowledge,
items or products in order to provide
recommendations. This filtering technique mark the
items for recommendations based on historical data
information, which is followed by the development
of an information recommender by using logical
reasoning technology (Burke, 2002). (Manouselis et
al., 2011; Zaina et al., 2011 and Zapata et al., 2013)
highlighted the need of RSs for TEL based on a
literature review which focuses on the availability
and ever increasing quantity of digital learning
resource repositories and from the outcomes of
Social Information Retrieval for Technology
Enhanced Learning (SIRTEL) annual workshop
series and a Special Issue on Social Information
Retrieval for TEL and proposed a DELPHOS and e-
LORS (e-learning object recommender system)
which are integral and intelligent solution for the
recommendation of learning objects (LO) stored in a
repository in which the recommendation are provided
in an ordered list of LOs’.
2.1.4 Hybrid Filtering
Hybrid RSs involve are the combination of
collaborative filtering and content-based filtering
techniques. (Adomavicius and Tuzhilin, 2005; Gu,
2013; Manouselis et al., 2012; Ricci et al., 2011)
(Ricci et al., 2011) suggest four different types of
hybrid recommenders: Separate collaborative and
content-based RSs; Collaborative RSs with added
features of content-based filtering method; Content-
based RSs with added features of collaborative
filtering method. (Poorni et al., 2014) presents a
personalized e-learning hybrid RS using the concept
of “Fuzzy Tree Matching” by considering the key
factors such as 1) learning activities and learners’
profiles 2) learning activities and 3) pedagogical
issues. (He et al., 2014) proposed A Social
Recommender System called as SRSH based on
Hadoop parallel computing platform. SRSH system
integrates content-based and collaborative filtering
techniques to further improve the performance of
recommendation. (Khribi et al., 2012) presents a
for building automatic recommendations
in e-learning platforms which consists of two
modules: an off-line module and an online module.
The first module pre-processes the data to build user
and content models. Online module uses these
models dynamically to find the user’s requirement
and goals, and predicts a list of recommendation.
User preferences objects are obtained by using a
“range of recommendation strategies” which are
primarily based on hybrid (content-based filtering
and collaborative filtering) filtering approach.
2.2 Personalized Learning RSs
Why personal? Every user, students in our case, has
individual needs and particular requirements. Some
of the students are highly self-motivated and learn by
exploring while other students prefer some specific
guidance in a structured way. With the ever growing
computer and internet technologies available, many
universities use LMSs’ to support teaching and
learning. Presently many LMSs’ are available as
proprietary solutions and open source (Santos and
Boticario, 2011). Despite the diversity in term of core
functionalities, these LMSs’ share a few
commonalities such as calendar, files, storage,
forums, wikis, blogs, etc. Education learning
systems, in particular LMS can further improve
teaching and learning practices of learners if
supported by personalized recommendations (Hauger
and Köck, 2007) and accessibility issues (Moreno et
al., 2009). The research finding in (Dagger et al.,
2007) shows that future LMSs’ primarily focus on
service-oriented architectures to incorporate social
media aspects into LMSs’.
In this section we present the proposed architecture
for a LMS recommendation system for
recommending learning material to students, which is
the primary objective of this paper. We selected
Moodle LMS in this case. The proposed architecture
is given in “Figure:1” and it consists of three main
components ‘learning material data source’, ‘seeking
student information’, and ‘generation’.
‘Learning material data source’ component is the
primary data source and this component is the
building block for designing and building our own
dataset for this research project. The dataset
developed here constitutes the knowledge base of
both formal and informal learners. The output of this
component is the preprocessed data which is
collected from:
MEC(Middle East College) Moodle Database -
former students competence qualifications;
Social Learners Network - wikis, forums, blogs
Other institutes LMS Servers. i.e., other three
institutes in Oman.
The recommendation from all the above can take
hybrid filtering technique though well-defined
educational metadata sources and educationally
influenced filtering decisions. The methodology
given in (Karampiperis and Diplaros, 2007) will be
applied, which generate a matrix to represent the
educational attributes of the learning resources. The
recommendation from all the above can take hybrid
filtering technique though well-defined educational
metadata sources and educationally influenced
filtering decisions. Additional filtering process can
be applied on this matrix based on an educational
“footprint” followed by the learners of these three
data sources.
The data acquired from MEC Moodle and LMS
of other institutes, connected using state of the art
“mutual authentication technology” which uses
Transport Layer Security protocol (Sheffer et at.,
2015), consist of formal learning data as well as
informal learning data acquired through social
learners’ networks.
Pre-processing of data allows the original data
from the above three sources to be transformed into a
Figure 1: Proposed architecture of a Personalized Recommendation System.
suitable form to be used by a data mining association
algorithm. The data pre-processing tasks in this
component includes data filtering, year and semester
wise session identification, student profiling, path
completion, transaction identification, data
transformation and enrichment, data integration, data
reduction to generate an explicit learning experience.
‘Seeking student information’ component
represents the various characteristics of the student,
including students profile creation, that can be used
to generate an implicit learning experience
using the
data mining association rules. This involves
understanding the “learning material attributes”,
which can be obtained from student’s current Moodle
session or previous logs of the student in social
learner network. Building student requirements
cannot be justified only by precise values, but also a
multi-criteria analysis approach need to be applied.
This is done by developing a multi criteria decision
model using linear weighted sum or a more rigorous
approach based on fuzzy set technology. The student
preferences and learning behavior can be defined
using the following three steps:
1. Clicks: This action defines the shortlisting of
2. Selection: This action is defined as material
selected and added to cart.
3. Learning: This action is using or reading the
The above three behaviors are used to identifying
the learners relative preferences LP
for each of the
material referred from the data sources
. The formula
used is:
 
 
 
denotes the number of references
to material through clicks, selection and learning
actions made by the learner i for material j
denotes the maximum
number of clicks, selection and learnings for a
learner i for M material.
denotes the
minimum number of clicks, selection and learnings
for a learner i for M material.
The first stage of the ‘Generation’ component is
to apply data mining rules and techniques on the
outputs of ‘learning material data source’ and
‘seeking student information’ components and
forward the results to a Hybrid RS. The objective of
this is to model students’ information seeking
behavior in order to develop a personalized
information retrieval system to enhance students’
learning experiences and such a PRS must have the
features given below. Students’ profile database has
the logs and records of learning styles, learning
material access and other profile of student with
respect to his/her specialization. These records and
logs will be used to identify the student’s preferences
with peers.
Some feature of educational RSs are suggested in
(López et al., 2015) and includes: monitoring
behavior, heuristics to infer information, user
feedback, filtering rules, navigation history, internal
database of items, similarity matching. The proposed
PRS will have the following features:
Good user interface design;
Real time recommendation as a service;
Accuracy of identifying student requirements;
Structured and accurate finding of recommended
learning materials;
Implicit and explicit recommendation of learning
In the last decade many researches have development
RSs for TEL but only a few of them validated these
RSs based on real time scenarios (Dascalu et al.,
2015). (Drachsler et al., 2010; Manouselis et al.,
2010) raises the issue of missing data sets for
recommender systems in TEL that can be used as
benchmarks to compare different recommendation
approaches. The availability of datasets helps in
drawing stronger conclusions about the validity and
generalizability of scientific experiments. This also
helps researchers to compare the experimental results
based on large datasets that capture learner
interactions in real settings. Furthermore, educational
datasets can support research advances on TEL
towards a basic theory for TEL (Verbert et al., 2011)
by offering the recorded and observed behaviour of
the stakeholders (students, teachers, parents, lifelong
learners, educational institutes) in different formal
and informal learning settings. (Verbert et al., 2011)
suggested guidelines for assembling suitable
datasets. We proposed a framework for building our
own dataset, “Figure: 2”, for the RS architecture
given in “Figure: 1”. The building of proposed
dataset framework is based on the guidelines given in
(Verbert et al., 2011) which include:
a. Create a data set that realistically reflects the
variables of the learning setting”;
b. “Use a sufficiently large set of user profiles”;
c. “Create data sets that are comparable to others”.
An evaluation of an interactive system ensures that it
behaves as expected by the designer and that it meets
the requirements of the user (Dix, 2009). (Ricci et al.,
2011) has provided some evaluation requirements of
TEL RSs and has proposed a four layer general
guidelines framework to evaluate the success of TEL
RSs: 1. Reaction of user, 2. Learning, 3. Behaviour,
4. Results. (Thai-Nghe et al., 2010) presented a
protocol for used for TEL RSs evaluation. The
evaluation algorithms were implemented in
MyMedia open source framework and the results
were compared with traditional methods such as
logistic regression or linear regression. (Drachsler et
al., 2009) highlighted the limitations that exists in the
evaluation of TEL RS due to unavailability of
datasets. They also suggested the framework given in
Table 1 for the analysis of TEL RSs.
Table 1: (adapted from (Drachsler et al., 2009)): An
evaluation framework for RSs in TEL.
Measurements Parameters
1. Accuracy
2. Performance
Educational measures
1. Effectiveness
2. Efficiency
3. Satisfaction
Social network measures
1. Variety
2. Centrality
In this paper we have presented a study of
evaluations of recommendation systems for TEL by
raising several concerns, issues and challenges
encountered by these systems. We also proposed a
personalized learning material recommendation
system architecture and discuss related technologies.
The proposed architecture has good
characteristics in recommending students to choose
appropriate learning materials for their assessments
Figure 2: Architecture building dataset for personalized material recommendation system.
by providing relevant recommendations. In the next
phase of this research project the framework will be
implemented as a plugin for the MEC Moodle to be
tested and validated, but not limited to, on the newly
developed dataset.
Aamir, M. and Bhusry, M., 2015. Recommendation
System: State of the Art Approach. International
Journal of Computer Applications, 120(12).
Adomavicius, G. and Tuzhilin, A., 2005. 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,
17(6), pp.734-749.
Burke, R., 2002. Hybrid recommender systems: Survey
and experiments. User modeling and user-adapted
interaction, 12(4), pp.331-370.
Bustos-López, M., Vásquez-Ramírez, R. and Alor-
Hernández, G., 2015. An Architecture for Developing
Educational Recommender Systems. Research in
Computing Science, 106, pp.17-26.
Candillier, L., Jack, K., Fessant, F. and Meyer, F., 2009.
State-of-the-art recommender systems. Collaborative
and Social Information Retrieval and
AccessTechniques for Improved User Modeling.
Dagger, D., O'Connor, A., Lawless, S., Walsh, E. and
Wade, V.P., 2007. Service-oriented e-learning
platforms: From monolithic systems to flexible
services. IEEE Internet Computing, 11(3).
Dascalu, M.I., Bodea, C.N., Moldoveanu, A., Mohora, A.,
Lytras, M. and de Pablos, P.O., 2015. A recommender
agent based on learning styles for better virtual
collaborative learning experiences. Computers in
Human Behavior, 45, pp.243-253.
del Castillo-Carrero, V., Hernán-Losada, I. and Martín, E.,
2014. Prototype of content-based recommender system
in an educational social network. In 1st Workshop on
Video based Learning, Graz, Austria. Retrieved from
org/lvcs/pdfs/paper/lvcs2014_submission_7. pdf.
Dix, A., 2009. Human-computer interaction. In
Encyclopedia of database systems (pp. 1327-1331).
Springer US.
Drachsler, H., Bogers, T., Vuorikari, R., Verbert, K.,
Duval, E., Manouselis, N., Beham, G., Lindstaedt, S.,
Stern, H., Friedrich, M. and Wolpers, M., 2010. Issues
and considerations regarding sharable data sets for
recommender systems in technology enhanced
learning. Procedia Computer Science, 1(2), pp.2849-
Drachsler, H., Hummel, H.G. and Koper, R., 2009.
Identifying the goal, user model and conditions of
recommender systems for formal and informal
learning. Journal of Digital Information, 10(2).
Fazeli, S., Loni, B., Drachsler, H. and Sloep, P., 2014,
September. Which recommender system can best fit
social learning platforms?. In European Conference on
Technology Enhanced Learning (pp. 84-97). Springer,
Fischer, G., 2001. User modeling in human–computer
interaction. User modeling and user-adapted
interaction, 11(1), pp.65-86.
Gu, Y.F., 2013. Research on personalized
recommendation system in e-learning based on
semantic web. In proceedings of the 3rd international
conference on multimedia technology (ICMT 2013)
(pp. 1046-1053).
Hauger, D. and Köck, M., 2007, September. State of the
Art of Adaptivity in E-Learning Platforms. In LWA
(pp. 355-360).
He, C., Tang, Y., Yang, Z., Zheng, K. and Chen, G., 2014.
SRSH: a social recommender system based on
Hadoop. International Journal of Multimedia and
Ubiquitous Engineering, 9(6), pp.141-152.
Hsu, M.H., 2008. A personalized English learning
recommender system for ESL students. Expert
Systems with Applications, 34(1), pp.683-688.
Introducing a powerful open source social networking
engine, 2017. Available from: http://elgg.org/. [15
May 2017].
Karampiperis, P. and Diplaros, A., 2007, September.
Exploiting Image Segmentation Techniques for Social
Filtering of Educational Content. In conference
Workshop on Social Information Retrieval for
Technology-Enhanced Learning (SIRTEL’07) at the
EC-TEL conference.
Karampiperis, P., Koukourikos, A. and Stoitsis, G., 2014.
Collaborative filtering recommendation of educational
content in social environments utilizing sentiment
analysis techniques. In Recommender Systems for
Technology Enhanced Learning (pp. 3-23). Springer
New York.
Khribi, M.K., Jemni, M. and Nasraoui, O., 2012.
Automatic personalization in e-learning based on
recommendation systems: An overview. In Intelligent
and Adaptive Learning Systems: Technology
Enhanced Support for Learners and Teachers (pp. 19-
33). IGI Global.
Kirkwood, A. and Price, L., 2014. 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), pp.6-
Kramár, T. and Bieliková, M., 2012, September.
Dynamically selecting an appropriate context type for
personalisation. In Proceedings of the sixth ACM
conference on Recommender systems (pp. 321-324).
Manouselis, N., Drachsler, H., Verbert, K. and Duval, E.,
2012. Recommender systems for learning. Springer
Science & Business Media.
Manouselis, N., Drachsler, H., Vuorikari, R., Hummel, H.
and Koper, R., 2011. Recommender systems in
technology enhanced learning. Recommender systems
handbook, pp.387-415.
Manouselis, N., Vuorikari, R. and Van Assche, F., 2010.
Collaborative recommendation of elearning
resources: an experimental investigation. Journal of
Computer Assisted Learning, 26(4), pp.227-242.
Moreno, L., Martínez Normand, L., Boticario, J.G. and
Fabregat, R., 2009. Research on standards supporting
A2UN@: Adaptation and Accessibility for ALL in
Higher Education.
Pazzani, M.J. and Billsus, D., 2007. Content-based
recommendation systems. In The adaptive web (pp.
325-341). Springer, Berlin, Heidelberg.
Poorni, G., Student, P.G., Balaji, K. and DeepthiNivetha,
C., A Personalized E-Learning Recommender System
Using the Concept of Fuzzy Tree Matching.
Prates, J.C., Fritzen, E., Siqueira, S.W., Braz, M.H.L. and
De Andrade, L.C., 2013. Contextual web searches in
Facebook using learning materials and discussion
messages. Computers in Human Behavior, 29(2),
Ricci, F., Rokach, L. and Shapira, B., 2011. Introduction
to recommender systems handbook. In Recommender
systems handbook (pp. 1-35). springer US.
Santos, O.C. and Boticario, J.G., 2011. Requirements for
semantic educational recommender systems in formal
e-learning scenarios. Algorithms, 4(2), pp.131-154.
Sheffer, Y., Holz, R. and Saint-Andre, P., 2015.
Recommendations for secure use of transport layer
security (TLS) and datagram transport layer security
(DTLS) (No. RFC 7525).
Shishehchi, S., Banihashem, S.Y., Zin, N.A.M. and Noah,
S.A.M., 2011, June. Review of personalized
recommendation techniques for learners in e-learning
systems. In Semantic Technology and Information
Retrieval (STAIR), 2011 International Conference
on (pp. 277-281). IEEE.
Thai-Nghe, N., Drumond, L., Krohn-Grimberghe, A. and
Schmidt-Thieme, L., 2010. Recommender system for
predicting student performance. Procedia Computer
Science, 1(2), pp.2811-2819.
Tzikopoulos, A., Manouselis, N. and Vuorikari, R., 2009.
An overview of learning object repositories. Selected
readings on database technologies and applications,
Verbert, K., Drachsler, H., Manouselis, N., Wolpers, M.,
Vuorikari, R. and Duval, E., 2011, February. Dataset-
driven research for improving recommender systems
for learning. In Proceedings of the 1st International
Conference on Learning Analytics and
Knowledge (pp. 44-53). ACM.
Verbert, K., Manouselis, N., Ochoa, X., Wolpers, M.,
Drachsler, H., Bosnic, I. and Duval, E., 2012. Context-
aware recommender systems for learning: a survey
and future challenges. IEEE Transactions on Learning
Technologies, 5(4), pp.318-335.
Zaina, L.A., Rodrigues Jr, J.F. and Bressan, G., 2010,
September. An approach to design the student
interaction based on the recommendation of e-learning
objects. In Proceedings of the 28th ACM International
Conference on Design of Communication (pp. 223-
228). ACM.
Zaina, L.A., Rodrigues Jr, J.F., de Andrade Cardieri,
M.A.C. and Bressan, G., 2011. Adaptive learning in
the educational e-LORS system: an approach based on
preference categories. International Journal of
Learning Technology, 6(4), pp.341-361.
Zapata, A., Menéndez, V.H., Prieto, M.E. and Romero, C.,
2013. A framework for recommendation in learning
object repositories: An example of application in civil
engineering. Advances in Engineering Software, 56,