; Adaptive E-Learning System based on Semantic Web and Fuzzy Clustering
Learning Center
Plans & pricing Sign in
Sign Out
Your Federal Quarterly Tax Payments are due April 15th Get Help Now >>

Adaptive E-Learning System based on Semantic Web and Fuzzy Clustering


The International Journal of Computer Science and Information Security (IJCSIS) is a well-established publication venue on novel research in computer science and information security. The year 2010 has been very eventful and encouraging for all IJCSIS authors/researchers and IJCSIS technical committee, as we see more and more interest in IJCSIS research publications. IJCSIS is now empowered by over thousands of academics, researchers, authors/reviewers/students and research organizations. Reaching this milestone would not have been possible without the support, feedback, and continuous engagement of our authors and reviewers. Field coverage includes: security infrastructures, network security: Internet security, content protection, cryptography, steganography and formal methods in information security; multimedia systems, software, information systems, intelligent systems, web services, data mining, wireless communication, networking and technologies, innovation technology and management. ( See monthly Call for Papers) We are grateful to our reviewers for providing valuable comments. IJCSIS December 2010 issue (Vol. 8, No. 9) has paper acceptance rate of nearly 35%. We wish everyone a successful scientific research year on 2011. Available at http://sites.google.com/site/ijcsis/ IJCSIS Vol. 8, No. 9, December 2010 Edition ISSN 1947-5500 � IJCSIS, USA.

More Info
  • pg 1
									                                                                    (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                       Vol. 8, No. 9, 2010 .

Adaptive E-Learning System based on Semantic Web and
                  Fuzzy Clustering
                       Khaled M. Fouad                                                                  Mofreh A. Hogo
Computer Science Dep. - Community College - Taif University,                  Computer Science Dep. – Computers and Information Systems
             Kingdom of Saudi Arabia (KSA);                                    College - Taif University, Kingdom of Saudi Arabia (KSA).
  Central Laboratory of Agriculture Expert Systems (CLAES), Egypt                              mofreh_hogo@yahoo.com

                     Shehab Gamalel-Din                                                                Nagdy M. Nagdy
    Computers Dept, Faculty of Science, King Abdul-Aziz                          Systems and Computers Engineering Dep. – Faculty of
       University, Kingdom of Saudi Arabia (KSA)                                      Engineering – AlAzhar University, Egypt.
                   smostafa@kau.edu.sa                                                        prof_nagdy@yahoo.com

Abstract— This work aims at developing an adaptive e-learning                A. ADAPTIVE E-LEARNING
system with high performance to reduce the challenges faces e-                   In the context of e-learning [2], adaptive e-learning systems
learners, the instructors and provides a good monitoring system              are more specialized and focus on the adaptation of learning
for the complete e-learning systems as well as the system                    content and the presentation of this content. According to [3],
structure. The work presents the different phases for the system             an adaptive system focuses on how the knowledge is learned
development of the adaptive system as: the first stage is the                by the student and pays attention to learning activities,
collection of the e-learners documents, the second stag is the               cognitive structures and the context of the learning material.
documents representation including the frequency count and the
weighting of the documents with its frequencies, the third stage is             The structure of an adaptive e-learning system is shown in
the prediction and clustering of e-learners interests using the              Fig. 1.
fuzzy clustering method and the statistical K-means clustering
                                                                                                                Data about user
method. The results obtained from this work shows that we have
to have different e-learners ontologies using the results of the                                                                              User Modeling
clustering methods which reflect the e-learners interests. Finally                 System

the work concluded the suggestions as well as the
recommendations for the instructors and the systems
                                                                                                                User Model
   Keywords-component; E-Learning; Semantic Web; Fuzzy
Ckustering; User model; User Model Representation.                                                              Adaptation

                                                                                            Fig. 1: The Structure of an Adaptive System [2]
                        I.     INTRODUCTION
                                                                                 The system intervenes at three stages during the process of
    Electronic learning or E-Learning [1] is interactive learning            adaptation. It controls the process of collecting data about the
in which the learning content is available online and provides               learner, the process of building up the learner model (learner
automatic feedback to the student's learning activities, it is               modeling) and during the adaptation process. It is not feasible
much like computer-based training (CBT) and computer-aided                   in conventional WBE to create static learning material [1] that
instruction (CAT), but the point is that it requires Internet for            can be read in any arbitrary sequence, because of many
access to learning material and for monitoring the student's                 interdependences and prerequisite relationships between the
activities. E-Learners usually can communicate with their                    course pages. However, adaptive hypermedia (AH) methods
tutors through the Internet.                                                 and techniques make it possible to inform learners that certain
    Semantic web could offer unprecedented support to the                    links lead to material they are not ready for, to suggest visiting
network teaching in semantic query, meaning construction,                    pages the learner should consult, or automatically provide
knowledge acquisition and sharing and collaborative learning.                additional explanations at the pages the learner visits, in order
Simultaneously semantic web also provides the support to                     to scaffold his/her progress. Adaptive educational hypermedia
describe semantics of learner characteristic in the learner                  systems (AEHSs) apply different forms of learner models to
model, and makes it possible to share learner model between                  adapt the content and the links of hypermedia course pages to
systems. So we need to construct the learner model and                       the learner. AEHSs support adaptive learning, using technology
clustering it; to simplify contents search that is based on that             to constantly measure the learner's knowledge and progress in
learner profile, in adaptive learning system based on semantic               order to adapt learning content delivery, presentation, feedback,
web.                                                                         assessment, or environment to the learner's needs, pace,
                                                                             preferences, and goals. Such systems make predictions of what
                                                                             the learner needs to attain his/her goals, respond to such needs,

                                                                       308                                   http://sites.google.com/site/ijcsis/
                                                                                                             ISSN 1947-5500
                                                              (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                 Vol. 8, No. 9, 2010 .
allocate resources, implement change, and thus improve                          The third layer accommodates RDF and RDF Schema as
personalization of the learning process. The system can be                 mechanisms to describe the resources available on the Web. As
designed to use predictive strategies prior to instruction                 such, they may be classified as lightweight ontology languages.
delivery and learning sessions, during the instruction (based on           Full ontology description languages appear in the fourth layer
the learner's interaction), or both.                                       as a way to capture more semantics. The topmost layer
    The open adaptive learning environment [4] is in which                 introduces expressive rule languages. The semantic web [7] is a
learners dynamically select a learning route suitable to their             space understandable and navigable by both human and
needs and profile. The proposed environment is based on the                software agents. It adds structured meaning and organization to
IEEE/IMS learning object metadata (LOM) standard. The                      the navigational data of the current web, based on formalized
nature of adaptations provided by this environment are centered            ontologies and controlled vocabularies with semantic links to
on the learner, and allow the LO to adapt to the evolving                  each other. From the E-Learning perspective, it aids learners in
learner’s model in terms of background, learning modalities,               locating, accessing, querying, processing, and assessing
and learning styles. It was summarized [1] the role of                     learning resources across a distributed heterogeneous network;
personalization in learning environments as follows:                       it also aids instructors in creating, locating, using, reusing,
  • Personalized learning environments enable one-to-one or                sharing and exchanging learning objects (data and
    many-to-one learning paradigms (one teacher - one learner,             components). The semantic web-based educational systems
    and many teachers – one learner), contrary to traditional              need to interoperate, collaborate and exchange content or re-
    learning environments that always adopt one-to-many                    use functionality. A key to enabling the interoperability is to
    learning paradigm (one teacher, many students);                        capitalize on (1) semantic conceptualization and ontologies, (2)
  • Personalized learning environments impose no constraints               common standardized communication syntax, and (3) large-
    in terms of learning time, location, etc., whereas traditional         scale service-based integration of educational content and
    ones are fairly restricted by the learning setting;
                                                                           functionality provision and usage. The vision of the semantic
  • Personalized learning environments recognize the huge                  web-based E-Learning is founded on the following major
    variety in the learner's characteristics and preferences in
    terms of the learning style, media, interests, and the like,
    and adapt instruction according to them; traditional ones are          • Machine-understandable educational content
    usually designed for the "average learner";                            • Shareable educational ontologies, including:
  • Personalized learning environments tailor instruction to suit                    Subject matter ontologies
    the learner's requirements (self-directed learning); in                          Authoring ontologies (modeling authors’ activities)
    traditional learning environments, the curriculum, learning            • Educational semantic web services, for supporting:
    units, and the selection and sequencing of learning material                     Learning, e.g., information retrieval, summarization,
    are determined by the tutor.                                                  interpretation (sense-making), structure-visualization,
                                                                                  argumentation, etc.
                                                                                     Assessment, e.g., tests and performance tracking
    It is generally agreed that credible evidence for mastery of                     Collaboration, e.g., group formation, peer help, etc.
learned material [5] is the goal of instructors. While educators           • Semantic interoperability
and domain experts agree that decoding meaning from text                       Semantic interoperability, the key promise of the semantic
plays a critical role in the acquisition of knowledge across all           web, is defined as a study of bridging differences between
disciplines, what particular evidence of mastery is required and           information systems on two levels as following:
what lends credibility to such evidence are the subjects of a
                                                                               • on an access level, where system and organizational
lively debate among experts in the learning community. The
need for new methods for semantic analysis of digital text is                    boundaries have to be crossed by creating standardized
now widely recognized in the face of the rising tide of                          interfaces that share system-internal services in a loosely-
information on the Web.         The layered model [6] for the                    coupled way; and
Semantic Web as shown in Fig. 2 puts the relationship among                   • On a meaning level, where agreements about transported
ontology description languages, RDF and RDF Schema, and                          data have to be made in order to permit their correct
XML in a better perspective. The bottom layer offers character                   interpretation. Interoperability requires the use of
encoding (Unicode) and referencing (URI) mechanisms. The                         standard SW languages for representing ontologies,
second layer introduces XML as the document exchange                             educational content, and services.
                                                                           C. THE ONTOLOGIES
                                                                               The word ontology comes from the Greek ontos, for being,
                                                                           and logos, for word. In philosophy, it refers to the subject of
                                                                           existence, i.e. to the study of being as such. More precisely, it
                                                                           is the study of the categories of things that exist or may exist
                                                                           in some domain [1]. Domain ontology explains the types of
                                                                           things in that domain. Ontology [8] comprises a set of
                                                                           knowledge terms, including the vocabulary, the semantic
         Fig. 2: The Architecture of the Semantic Web                      interconnections, and some simple rules of inference and logic

                                                                     309                             http://sites.google.com/site/ijcsis/
                                                                                                     ISSN 1947-5500
                                                             (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                Vol. 8, No. 9, 2010 .
for some particular topic. Ontologies applied to the Web are              hand the LIP standard is based on the classical notion of a CV
creating the Semantic Web. Ontologies provide the necessary               and inter-personal relationships are not considered at all.
armature around which knowledge bases should be built, and
set grounds for developing reusable Web-contents, Web-                    E. K-MEANS CLUSTERING METHOD
services, and applications. Ontologies facilitate knowledge                   Clustering of objects is as ancient as the human need for
sharing and reuse, i.e. a common understanding of various                 describing the salient characteristics of men and objects and
contents that reaches across people and applications.                     identifying them with a type. Therefore, it embraces various
Technically, an ontology is a text-based piece of reference-              scientific disciplines: from mathematics and statistics to
knowledge, put somewhere on the Web for agents to consult it              biology and genetics, each of which uses different terms to
when necessary, and represented using the syntax of an                    describe the topologies formed using this analysis.
ontology representation language. There are several such                      The simplest and most commonly used algorithm,
languages around for representing ontologies, for an overview             employing a squared error criterion is the K-means algorithm
and comparison of them. It is important to understand that                [23]. This algorithm partitions the data into K clusters (C1,C2,
most of them are built on top of XML and RDF.                             . . . ,CK), represented by their centers or means. The center of
    The most popular higher-level ontology representation                 each cluster is calculated as the mean of all the instances
languages were OIL (Ontology Inference Layer) and
                                                                          belonging to that cluster. The algorithm [23] starts with an
DAML+OIL. An ontology developed in any such language is
                                                                          initial set of cluster centers, chosen at random or according to
usually converted into an RDF/XML-like form and can be
partially parsed even by common RDF/XML parsers. Of                       some heuristic procedure. In each iteration, each instance is
course, language-specific parsers are necessary for full-scale            assigned to its nearest cluster center according to the
parsing. There is a methodology for converting an ontology                Euclidean distance between the two. Then the cluster centers
developed in a higher-level language into RDF or RDFS. In                 are re-calculated. The center of each cluster is calculated as the
early 2004, W3C has officially released OWL (Web Ontology                 mean of all the instances belonging to that cluster:
Language) as W3C Recommendation for representing
ontologies. OWL is developed starting from description logic
and DAML+OIL. The increasing popularity of OWL might                          Where N k is the number of instances belonging to cluster k
lead to its widest adoption as the standard ontology
representation language on the Semantic Web in the future.                and µ is the mean of the cluster k.
Essentially, OWL is a set of XML elements and attributes, with                A number of convergence conditions are possible. For
well-defined meaning, that are used to define terms and their             example, the search may stop when the partitioning error is
relationships (e.g., Class, equivalent Property, intersection Of,         not reduced by the relocation of the centers. This indicates that
union Of, etc.).                                                          the present partition is locally optimal. Other stopping criteria
                                                                          can be used also such as exceeding a pre-defined number of
D. LEARNER MODEL AND PROFILE                                              iterations. Figure 3 presents the pseudo-code [23] of the K-
    The behavior of an adaptive system [9] varies according to            means algorithm.
the data from the learner model and the learner profile. Without              Input: S (instance set), K (number of cluster)
knowing anything about the learner, a system would perform in                 Output: clusters
exactly the same way for all learners. It was described the                   1: Initialize K cluster centers.
application of learner models as follows:                                     2: while termination condition is not satisfied do
    An extensive learner model must contain information about                 3: Assign instances to the closest cluster center.
the learner’s domain knowledge, the learner’s progress,                       4: Update cluster centers based on the assignment.
preferences, goals, interests and other information about the
                                                                              5: end while
learner, which is important for the used systems. Learner                                     Fig. 3. K-means Algorithm.
models can be classified according to the nature and form of
information contained in the models. Considering the subject                 The rest of the paper is organized as following: Section 2 is
domain, the information stored in a learner model can be                  reserved for the related works; section 3 introduces the
divided into two major groups: domain specific information                proposed system structure architecture, section 4 presents the
and domain independent information.                                       experiments design and results analysis, section 5 presents the
    We examined two of the most important and well-                       concluded suggestions and recommendations to improve the
developed standards - the PAPI standard [10] and the IMS LIP              system performance, finally section 6 concludes the work and
standard [11]. Both standards deal with several categories for            introduces the future work.
information about a learner. These standards have been
developed from different points of view. The PAPI standard                                   II.   RELATED WORKS
reflects ideas from intelligent tutoring systems where the                   An accurate representation of a learners interests [12],
performance information is considered as the most important               generally stored in some form of learner model, is crucial to
information about a learner. The PAPI standard also stresses on           the performance of personalized search or browsing agents.
the importance of inter-personal relationships. On the other              Learner model is often represented by keyword/concept
                                                                          vectors or concept hierarchy. The acquired model can then be

                                                                    310                             http://sites.google.com/site/ijcsis/
                                                                                                    ISSN 1947-5500
                                                            (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                               Vol. 8, No. 9, 2010 .
used for analyzing and predicting the future learner access              completes the user profiles model consistently on the related
behavior. Learner model may be built explicitly, by asking               feedback mechanism.
learners questions, or implicitly, by observing their activity.              In [16], the authors proposed a new approach to User
   In [13], authors investigated the techniques to create a user         Model Acquisition (UMA) which has two important features. It
profile automatically using the ontological approach. They               doesn’t assume that users always have a well-defined idea of
used a framework to gather the user information from different           what they are looking for, and it is ontology-based, i.e., it was
search space where user’s details could be found. The details            dealt with concepts instead of keywords to formulate queries.
include user’s general information to specific preferences.              The first problem is that most approaches assume users to have
They used Meta search in user’s blog, personal/organization              a well-defined idea of what they are looking for, which is not
web page, and any other cites to collect information about               always the case. They solved this problem by letting fuzzy user
user. This information is assigned to a pre- structure hierarchy         models evolve on the basis of a rating induced by user
or in reference ontology to create an initial user profile. More         behavior. The second problem concerns the use of keywords,
clearly, initial profile is learned by the concept/document              not concepts, to formulate queries. Considering words and not
collected from user’s details. In traditional user profiling             the concepts behind them often leads to a loss in terms of the
system feature extraction from document is done by vector                quantity and quality of information retrieved. They solved this
space model or considering term frequency, tf-idf methods                problem by adopting an ontology-based approach.
only. In this research, authors considered WordNet and                       In [9], the student interactions with the system are
Lexico-Syntactic pattern for hyponyms to extract feature from            monitored and stored in log files. The recorded data are then
document. This profile further improved by taking                        cleaned and preprocessed (e.g. compute the relative frequency
collaborative user methods. Where, they found a group of                 of learner actions, the amount of time spent on a specific action
users with similar interest by taking similarity score among             type, the order of navigation etc). Subsequently, these
them. After that an ontology matching approach is applied to             behavioral indicators are analyzed and based on them the
learn the profile with other similar user which is called                system can infer different learning preferences of the student.
improved profile. [3] Introduced a method for learning and               Finally, the identified learner model is used by the decision-
updating a user profile automatically. The proposed method               making component to select the most appropriate adaptation
belongs to implicit techniques – it processes and analyzes               actions, in order to provide the student with the educational
behavioral patterns of user activities on the web, and modifies          resources that suit her/his specific needs.
a user profile based on extracted information from user’s web-
logs. The method relies on analysis of web-logs for                                    III.      THE PROPOSED SYSTEM ARCHITECTURE
discovering concepts and items representing user’s current and               The design idea of adaptive learning system based on
new interests. Those found concepts and items are compared               creating an ideal learning environment for the learners so that
with items from a user profile, and the most relevant ones are           system can provide the adaptive learning support according to
added to this profile. The mechanism used for identifying                the learner's individual differences, and to promote learners to
relevant items is built based on a newly introduced concept of           study initiatively, and to achieve the knowledge construction.
ontology-based semantic similarity. There are many different             There are some objectives in the design of adaptive learning
methods to Construct Learner Models [2]: Machine Learning                system. First, system can provide the adaptive learning content
Methods, Bayesian Methods, Overlay methods, Stereotype                   based on the learner's interest and knowledge requirement.
methods, Plan Recognition. Update of Learner Models                      Second, system can support the self-directed learning and
methods are: Analysis of Learner Responses, Analysis of the              collaborative Learning. Third, system can help teachers
                                                                         understand the learning process of learners, and adjust the
Process of Problem Solution, Analysis of Learner Actions,
                                                                         pedagogical activities, and support the learning evaluation.
Discounting Old Data.                                                    Last, system will support the courses development for staff.
    In [14], student model mainly included the cognitive model           Based on these considerations, a new architecture of adaptive
and the interest model. Cognitive model mainly pay attention to          learning system is proposed in current paper. It is illustrated in
learner background knowledge, study style and cognition level.           figure 4. According to proposed architecture, our learning
Through synthesizing the domestic and foreign research                   system is mainly composed of four processes: Learner's Web
practice authors proposed to use Solomon Study Style Measure             Log Analysis, Learner Interest Builder, and Knowledge
Meter as preceding measure to test learner's study style.                Requirement Acquiring. These Processes will be explained in
Regarding the student’s cognition level’s estimate they took the         the next section.
thought of fuzzy set. It was combined the ontology and concept
space [15], indicated the feature items of user profile with              Learner
                                                                                                                          Visited pages
                                                                                                                           & Learner             Interest
semantic concepts, calculates learner’s interest-level to the                                                               Behavior            Acquiring

topic through establishing the word frequency and utilize the
                                                                                                     Learner profile
suitable calculation methods, mining the concepts within the                                             cluster
user’s feedback files and the relationship between concepts,                                                                                    Domain
combines user’s short-term interests and long-term interests to              Representation            Generating            Learner Interest
create user profiles model with semantic concept hierarchy tree                of profile
                                                                                                     Representation of

and embody the drifting of user profile and improves and

                                                                                                     Fig.4. The Architecture of the Proposed

                                                                   311                                              http://sites.google.com/site/ijcsis/
                                                                                                                    ISSN 1947-5500
                                                                            (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                               Vol. 8, No. 9, 2010 .
  A.     LEARNER'S INTEREST ACQUIRING                                                  •    DOMAIN CONCEPT FILTERING
    In this system, learner interest model’s knowledge                               This process discovers concepts which represent the learner’s
expression uses the thought which is based on the space vector                       interests. These concepts and items are compared to the
model’s expression method and the domain ontology [22]. The                          domain ontology to check the relevant items to the learner
figure 5 shows certain steps to acquire learner interest.                            profile. The most relevant ones update the learner profile. The
   Web -
                   Document         Extracted Concepts
                                                                                     items relevance is based on ontology-based semantic
                                                                                     similarity where browsed items by a learner on the web are
                                                                                     compared to the items from a domain ontology and learner
                                                                                     profile. The importance is combined with the semantic
                                   Domain Concept          Domain-Related
                                                                                     similarity to obtain a level of relevance. The page items are
                                                                                     processed to identify domain-related words to be added to the
                  Learner              Domain
                                                                                     learner profile. A bag of browsed items is obtained via a
                   Model               Ontology                                      simple word indexing of the page visited by the learner. We
                                                                                     filter out irrelevant words using the list of items extracted from
                                                                                     domain ontology. Once domain-related items are identified,
                                                                                     we evaluate their relevance to learner’s interests. The selected
              Fig.5: Acquiring E-learners interest Steps                             method was used in [20, 21] to compute semantic similarity
                                                                                     function (S) based on a domain ontology. The similarity is
  •      DOCUMENT REPRESENTATION                                                     estimated for each pair of items where one item is taken from
    The Vector Space Model [17,18, 19] is adapted in our                             a learner profile, while the other one from a set of browsed
proposed system to achieve effective representations of                              items. The functions Sw is the similarity between synonym
documents. Each document is identified by n-dimensional                              sets, Su is the similarity between features, and Sn is the
feature vector where each dimension corresponds to a distinct                        similarity between semantic neighborhoods between entity
term. Each term in a given document vector has an associated
                                                                                     classes a of ontology p and b of ontology q, and ww , wu , and
weight. The weight is a function of the term frequency,
collection frequency and normalization factors. Different                            wn are the respective weights of the similarity of each
weighting approaches may be applied by varying this function.
                                                                                     specification component.
Hence, a document j is represented by the document vector dj:
      d j = ( w1 j , w2 j ,..., wnj ) Where, wkj is the weight of the                S(ap , bq ) = ww × Sw(ap , bq ) + wu × Su (ap , bq ) + wn × Sn (ap , bq ) ;

kth term in the document j.                                                          For ww ; wu ; wn ≥ 0:
    The term frequency reflects the importance of term k within
a particular document j. The weighting factor may be global or                       Weights assigned to Sw, Su, and Sn depends on the
local. The global weighting factor takes into account the                            characteristics of the ontologies.
importance of a term k within the entire collection of                                    The similarity measures are defined in terms of a
documents, whereas a local weighting factor considers the                            matching process [20, 21]:
given document only. Document keywords were extracted by                                                              AI B
using a term-frequency-inverse-document-frequency (tf-idf)                            S(a, b) =
calculation [18, 19], which is a well-established technique in                                     A I B + α ( a, b) A / B + (1 − α ( a, b)) B / A
information retrieval. The weight of term k in document j is                         where A and B are description sets of classes a and b, i.e.,
represented as:                                                                      synonym sets, sets of distinguishing features and a set of
                                                                                     classes in semantic neighborhood; (A∩B) and (A/B) represent
                 wkj = tf kj × (log n − log dfk + 1)
                                    2       2                                        intersection and difference respectively, | | is the cardinality of
                                                                                     a set; and α is a function that defines relative importance of
Where: tfkj is the term k frequency in document j, dfk is                            non-common characteristics. A set of browsed items that are
number of documents in which term k occurs, n = total number                         similar to items from the user profile is considered as a set of
of documents in collection.                                                          items that can be added to this profile. Table 2 shows a
    Table 1 shows the term frequency in different documents.                         sample of the weighted terms in the documents; that
The main purpose of this step is to extract interested items in                      found in table 1.
the web page, then get term frequency that reflects the
importance of term, finally get the weight of terms in the
selected page. The output of this step is the weight of terms in
selected page that can be used to build learner interest profile.

                                                                               312                                 http://sites.google.com/site/ijcsis/
                                                                                                                   ISSN 1947-5500
                                                                                    (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                                       Vol. 8, No. 9, 2010 .
                                                            Table 1: Sample of the Documents with their representation
                              Computer                                                                                            Computer                   Parallel
                               science        AI           Programming Software eng.    Networks          LAN          WAN          Arch.     Processors   processing
              Doc1              20           25                20              15             10           0            0               5          0            0
              Doc 2             15           15                25              15              5           0            0               0          0            0
              Doc 3              0            5                25              10             25           10           15              5          0            0
              Doc 4              0            5                25              10             20           5            15              5          0            0
              Doc 5              0            0                 5              0              10           0            0               5         20           20
              Doc 6              0            0                 0              5              20           0            0               0         25           30
              Doc 7              0            0                 5              0               0           0            0               5         20           10
              Doc 8              0            0                 0              5               0           0            0               10        25            5
              Doc 9             15            5                30              30              0           0            0               5          0            0
              Doc 10             5            0                25              25              0           0            0               0          0            0
              Doc 11            10            0                 0              0              10           0            0               25        10           30
              Doc 12            10            0                 0              0              10           0            0               30        10           25
              Doc 13            20           25                 5              0              10           0            0               5          0            0
              Doc 14            15           15                 5              0               5           0            0               0          0            0
              Doc 15             0           30                20              15             25           0            5               0          0            0
              Doc 16             0           25                25              15             20           0            0               0          0            0

                                                           Table 2: Sample of the weighted terms in the documents

                  computer                                        Software                                                    Computer                       Parallel
                                      AI     programming                             Network            LAN     WAN                          Processors
                   Science                                       Engineering                                                 Architecture                  Processing

       Doc1           40.00          45.75         28.30            25.17              14.15            0.00    0.00            9.15            0.00          0.00

       Doc2           30.00          27.45         35.38            25.17              7.08             0.00    0.00            0.00            0.00          0.00

       Doc3            0.00          9.15          35.38            16.78              35.38            40.00   51.23           9.15            0.00          0.00

       Doc4            0.00          9.15          35.38            16.78              28.30            20.00   51.23           0.00            0.00          0.00

       Doc5            0.00          0.00          7.08             0.00               14.15            0.00    0.00            9.15           48.30         48.30

       Doc6            0.00          0.00          0.00             8.39               28.30            0.00    0.00            0.00           60.38         72.45

       Doc7            0.00          0.00          7.08             0.00               0.00             0.00    0.00            9.15           48.30         24.15

       Doc8            0.00          0.00          0.00             8.39               0.00             0.00    0.00            18.30          60.38         12.08

       Doc9           20.00          9.15          42.45            50.34              0.00             0.00    0.00            9.15            0.00          0.00

      Doc10           10.00          0.00          35.38            41.95              0.00             0.00    0.00            0.00            0.00          0.00

      Doc11           20.00          0.00          0.00             0.00               14.15            0.00    0.00            45.75          24.15         72.45

      Doc12           20.00          0.00          0.00             0.00               14.15            0.00    0.00            54.90          24.15         60.38

      Doc13           40.00          45.75         7.08             0.00               14.15            0.00    0.00            9.15            0.00          0.00

      Doc14           30.00          27.45         7.08             0.00               7.08             0.00    0.00            0.00            0.00          0.00

      Doc15            0.00          54.90         28.30            25.17              35.38            0.00    17.08           0.00            0.00          0.00

      Doc16            0.00          45.75         35.38            25.17              28.30            0.00    0.00            0.00            0.00          0.00

                                                                                                         number of instances was 16, the number of attributes
                                                                                                         was 10, the number of clusters was 2 and their centers
IV.    THE EXPERIMENTS DESIGN AND RESULTS ANALYSIS                                                       are shown in Table 4
1.    The First Experiment: predicting the e-learners                                              3.    The Third Experiment: predicting the e-learners
      interests using the Simple Fuzzy-KMeans Where the                                                  interests using the fuzzy K-Means and attributes
      number of instances was 16, the number of attributes                                               selection using weka.attributeSelection.BestFirst.
      was 10, the number of clusters was 2 and their centers
      are shown in Table 3.
2.    The Second Experiment: predicting the e-learners
      interests using the weka.clusterers.XMeans, Where the

                                                                                        313                                       http://sites.google.com/site/ijcsis/
                                                                                                                                  ISSN 1947-5500
                                                                                        (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                                           Vol. 8, No. 9, 2010 .

                                                                          Table 3: Results of the first experiment
                                                         Computer                                   Software                                          Computer                         Parallel
                    The learners group
                                                          science         AI      Programming         eng.       Networks     LAN         WAN           Arch.      Processors        processing
      Cluster 0 number of cases belong to it are 10 (
                                                              9           15           20.5          13.5              12        1.5      3.5           2.5              0               0
       Cluster 1 number of cases belong to it are 6 (
                                                              3.3         0            1.66          1.66          8.33           0           0         12.5        18.33               20

                                                                        Table 3: Results of the Second experiment
                The learners group                Computer                                     Software                                             Computer                       Parallel
                                                   science          -AI       Programming        eng.          Networks     LAN        WAN            Arch.      Processors      processing
       Cluster 0 number of cases belong to it
                                                        7.8         10           11.7              8.9           8.9         0         0.35           6.4           7.8                8.5
                    are 14 ( 88%)
       Cluster 1 number of cases belong to it
                                                        0           5             25               10           22.5        7.5          15            5             0                  0
                     are 2 ( 13%)

                                            Table 4: Results of the Third experiment (The number of selected attributes are: 5)
                                    The learners group                                        AI          Programming
                                                                                                                               Arch.                                         processing
            Cluster 0 number of cases belong to it are 10 cases 63%                           15               20.5               2.5                       0                    0
             Cluster 1 number of cases belong to it are 6 cases 38%                           0                1.666              12.5                  18.333                  20

                                                                                                    correspond to the several ontologies. Ontology-based user
          V.      SUGGESTIONS AND RECOMMENDATIONS                                                   profiles typically maintain sophisticated representations of
                                                                                                    personal interest profiles. These representations can be utilized
   The Semantic Web can be used to organize information in                                          for effective information retrieval in the e-learning systems.
concept structures, while web services allow the encapsulation
of heterogeneous knowledge and modularization of the
architecture. The key idea of the Semantic Web is to have data                                                                           VI.         CONCLUSION
defined and linked in such a way that its meaning is explicitly
interpretable by software processes rather than just being                                             In this paper, we presented a method for constructing
implicitly interpretable by humans.                                                                 learner model that represents the user’s interests by analyzing
   The user profiles can maintain sophisticated representations                                     the web-log to extract the interested terms in visited pages by
of personal interest profiles. These representations can be                                         learner. Then the fuzzy clustering is used to extract clusters of
utilized for effective information retrieval. Fuzzy clustering                                      output learner profiles The goal of incorporating the semantic
allows an entity to belong to more than one cluster with                                            web is to build the semantically enhanced user models. Fuzzy
different degrees of accuracy, while hard clustering assigns                                        technique is used to cluster the extracted data of learner model
each entity exactly to one of the clusters. Thus, fuzzy                                             to classifying the learners for their interests. We recommend
clustering is suitable in constructing the learner profiles                                         to use Ontology-based user profiles to maintain sophisticated
representations such as (learner ontology).Such representation                                      representations of personal interest profiles.
of user profiles is useful because some information is not
forced to fully belong to any one of the user profiles. Fuzzy
clustering methods may allow some information to belong to                                                                             ACKNOWLEDGMENT
several user profiles simultaneously with different degrees of                                         The authors wish to thank Taif University for the support
accuracy.                                                                                           of this research and the anonymous reviewers for their
   Our proposed approach depends on using semantic web to                                           valuable comments.
extract the learner model. Fuzzy technique is used to cluster
the extracted data of learner model to classifying the learners                                                                               REFERENCES
for their interests. This classifying the interests of learners                                     [1]     D. Vladan. (2006), SEMANTIC WEB AND EDUCATION, ISBN: 0-
enable learning systems to handle learner as groups to                                                      387-35417-4, Springer Science+Business Media, LLC.
recommend them what they must teach corresponding to their                                          [2]     F. Christoph, (2005). User Modeling and User Profiling in Adaptive E-
                                                                                                            learning Systems, Master’s Thesis At Graz University of Technology.
interests in the learner profile clusters.
                                                                                                    [3]     R. Marek and K. Sayed. (2009), Updating User Profile using Ontology-
   Cluster of learner profiles must be used in e-learning                                                   based Semantic Similarity, FUZZ_IEEE 2009, Korea, August 20-24,
system, so we recommend to represent the cluster of profiles                                                978-1-4244-3597-5, IEEE.
on learner ontology to express sophisticated user profiles in                                       [4]     A. YACINE, B. RACHID, B. JAWAD. (2004), Learning Objects Based
order that a user may own several interests. A user profile may                                             Framework for Self-Adaptive Learning, Education and Information

                                                                                              314                                                 http://sites.google.com/site/ijcsis/
                                                                                                                                                  ISSN 1947-5500
                                                                         (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                            Vol. 8, No. 9, 2010 .
       Technologies 8:4, 345–368, 2003, 2004 Kluwer Academic Publishers.              [20] R. Marek and K. Sayed. (2009), Updating User Profile using Ontology-
       Manufactured in the Netherlands.                                                    based Semantic Similarity, FUZZ_IEEE 2009, Korea, August 20-24,
[5]    S. Uri, E. Masha. (2006), e-Learning for depth in the Semantic Web,                 978-1-4244-3597-5, IEEE.
       British Journal of Educational Technology, Vol 37 No 3 2006 425–444,           [21] M. Andrea, E. Max J. (2003), Determining Semantic Similarity among
       Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4                 Entity Classes from Different Ontologies, IEEE TRANSACTIONS ON
       2DQ, UK and 350 Main Street, Malden, MA 02148, USA.                                 KNOWLEDGE AND DATA ENGINEERING, VOL. 15, NO. 2,
[6]    K. Breitman, M.Casanova and W. Truszkowski. (2007), Semantic Web:                   MARCH/APRIL, 1041-4347/03, IEEE.
       Concepts, Technologies and Applications, e-ISBN: 978-1-84628-710-7,            [22] H. Hany, F. Kh. (2010), “Semantic web based Approach to learn and
       Springer-Verlag London Limited.                                                     update Learner Profile in Adaptive E-Learning ”, Al-Azhar Engineering
[7]    D. Dicheva. (2008), Ontologies and Semantic Web for E-Learning, in                  Eleventh International Conference, December 23-26, pp., 2010.
       book "Handbook on Information Technologies for Education and                   [23] M. Oded, R. Lior. (2010) , Data Mining and Knowledge Discovery
       Training", 978-3-540-74155-8 (Online), Springer Berlin Heidelberg.                  Handbook Second Edition, In chapter " A survey of Clustering
[8]    G. Fayed, D. Sameh, H. Ahmad, M. Jihad., A. Samir, and S. Hosam.                    Algorithms " ISBN 978-0-387-09822-7, Springer Science+Business
       (2006), E-Learning Model Based On Semantic Web Technology,                          Media.
       International Journal of Computing &Information Sciences Vol. 4, No.
       2, August 2006, On-Line. Pages 63 – 71.
[9]    P. Elvira, T. Philippe, B. Costin.(2007), Adaptive Educational
       Hypermedia Systems: A Focus on Learning Styles, EUROCON 2007
                                                                                                                 AUTHORS PROFILE
       The International Conference on “Computer as a Tool” Warsaw,
       September 9-12, 1-4244-0813-X/07, IEEE.                                                         Khaled M. Fouad received his Master degree of AI and
                                                                                                       expert systems. He is currently a PhD candidate in the faculty
[10]   LTSC Learner Model Working Group of the IEEE (2000), IEEE                                       of engineering AlAzhar University in Egypt. He working
       p1484.2/d7, 2000-11-28 Draft Standard for Learning Technology -                                 now as lecturer in Taif university in Saudi Arabia and is
       Public and Private Information (PAPI) for Learners (PAPI Learner),                              assistant researcher in Central Laboratory of Agriculture
       Technical              report             Available            Online:                          Expert Systems (CLAES) in Egypt. His current research
       http://www.edutool.com/papi/papi_learner_07_main.pdf                                            interests focus on Semantic Web and Expert Systems.
[11]   Smythe, C., F. Tansey, R. Robson (2001), IMS Learner Information
       Package Information Model Specification, Technical report Available                               Mofreh A. Hogo is a assistant professor at Benha
       Online: http://www.imsglobal.org/profiles/lipinfo01.html.                                         University, Egypt. He is a assistant professor of
                                                                                                         Computer Science and Engineering. Dr. Hogo holds a
[12]   12. [12] F. Christoph. (2005), User Modeling and User Profiling in                                PhD in Informatics Technology from Czech Technical
       Adaptive E-learning Systems, Master’s Thesis at Graz University of                                University in Prague, Computer Science and Engineering
       Technology, Copyright 2005 by Christoph Fr¨oschl.                                                 Dept. 2004. He is the author of over 40 papers that have
[13]   D. Trong, N Mohammed, L. Delong , and J. Geun. (2009), A                                          been published in refereed international Journals
       Collaborative Ontology-Based User Profiles System, N.T. Nguyen, R.                                (Information Sciences, Elsiver, UBICC, IJICIS, IJCSIS,
       Kowalczyk, and S.-M. Chen (Eds.): ICCCI 2009, LNAI 5796, pp. 540–                                 IJPRAI, ESWA, IJEL, Web Intelligence and Agent
       552, Springer-Verlag Berlin Heidelberg.                                                           Systems, Intelligent Systems, international journal of
                                                                                                         NNW, IJAIT Journal of Artificial Intelligence Tools,
[14]   B. Qiu, W. Zhao. (2009), Student Model in Adaptive Learning System                                IJCI).
       based on Semantic Web, In First International Workshop on Education
       Technology and Computer Science, 978-0-7695-3557-9/09, IEEE, DOI
       10.1109/ETCS.2009.466.                                                                          Shehab Gamalel-Din is Full Professor in King Abdul-Aziz
                                                                                                       University, Faculty of Science, Computers Dept. Jeddah, KSA
[15]   W. Cuncun, H. Chongben, T. Hengsong.(2009), A Personalized Model                                and Al-Azhar University, Department of Computers and
       for Ontology-driven User Profiles Mining, 2009 International                                    Systems Engineering Cairo, Egypt. He has Professional
       Symposium on Intelligent Ubiquitous Computing and Education, 978-0-                             Memberships in Institute of Electrical and Electronics Engineers
       7695-3619-4/09, IEEE.                                                                           (IEEE), USA, Association of Computing Machinery (ACM),
[16]   C´elia and T. Andrea G.B. (2006), An Evolutionary Approach to                                   USA, and Egyptian Society for Engineers, Egypt.
       Ontology-Based User Model Acquisition, V. Di Ges´u, F. Masulli, and
       A. Petrosino (Eds.): WILF 2003, LNAI 2955, pp. 25–32, c_Springer-
       Verlag Berlin Heidelberg.                                                                        Nagdy Mahmoud Nagdy is Professor of engineering
[17]   B. Qiu, W. Zhao. (2009), Student Model in Adaptive Learning System                               applications and computer systems, Department of
       based on Semantic Web, 2009 First International Workshop on                                      Systems Engineering and Computer Engineering -
       Education Technology and Computer Science, 978-0-7695-3557-9/09,                                 Faculty of Engineering Al-Azhar University. He
       IEEE, DOI 10.1109/ETCS.2009.466                                                                  received his PH.D in 1986. He has Supervision of some
                                                                                                        master's and doctoral degrees in the Department of
[18]   P. Jianguo, Z. Bofeng, W. Shufeng, W. Gengfeng, and W. Daming.
                                                                                                        Systems Engineering and Computer and Electrical
       (2007), Ontology Based User Profiling in Personalized Information
       Service Agent, Seventh International Conference on Computer and
       Information Technology, 0-7695-2983-6/07, IEEE.
[19]   P. Jianguo, Z. Bofeng, W. Shufeng, W. Gengfeng. (2007), A
       Personalized Semantic Search Method for Intelligent e-Learning, 2007
       International Conference on Intelligent Pervasive Computing, 0-7695-
       3006-0/07, IEEE , DOI 10.1109/IPC.2007.48.

                                                                                315                                  http://sites.google.com/site/ijcsis/
                                                                                                                     ISSN 1947-5500

To top