Data mining for adaptive learning sequence in
English language instruction
Ya-huei Wang , Ming-Hseng
Tseng , Hung-Chang Liao
Abstract[/B]
The purpose of this paper is to propose an
adaptive system analysis for optimizing learning sequences. The
analysis employs a decision tree algorithm, based on students’
profiles, to discover the most adaptive learning sequences for a
particular teaching content. The profiles were created on the basis
of pretesting and posttesting, and from a set of five student
characteristics: gender, personality type, cognitive style,
learning style, and the students’ grades from the previous
semester. This paper address the problem of adhering to a fixed
learning sequence in the traditional method of teaching English,
and recommend a rule for setting up an optimal learning sequence
for facilitating students’ learning processes and for max-imizing
their learning outcome. By using the technique proposed in this
paper, teachers will be able both to lower the cost of teaching and
to achieve an optimally adaptive learning sequence for students.
The results show that the power of the adaptive learning sequence
lies in the way it takes into account stu-dents’ personal
characteristics and performance; for this reason, it constitutes an
important innovation in the field of Teaching English as a Second
Language (TESL).
1. [/B]
Introduction[/B]
In order to compete and survive in the
twenty-first century glo-bal economy, it is essential that students
acquire communication skills in English ( Chen, Warden, &
Chang, 2006 ). The goal of teach-ing English – including
comprehension, listening, speaking, read-ing, and writing
proficiency – is to facilitate students’ future academic and
professional careers. Students’ learning depends on what happens in
the classroom, and in different classrooms there may be different
cognitive and learning styles. In the conventional learning systems
of Taiwan, however, teachers of English teach the same content to
all students, without taking into consideration the individual
students’ gender, personality type, cognitive style, learning
style, or previous knowledge. That is, the current courses are
based on ‘‘static” learning material, not ‘‘dynamic” learning
material (Romero, Ventura, Delgado, & Bra, 2007). In this type
of learning system, if students wish to maximize their learning
out-come, they must adapt themselves to the course content, the
course content is never adapted to accommodate their individual
needs and preferences.
Adapting what goes on in the classroom to
students’ needs in-volves two important issues: how to tailor
courses to each individual students’ characteristics and
capabilities, and how to create, repre-sent, and maintain the
activity tree with the appropriate associated sequencing definition
for different students. Unfortunately, because of the enormous
costs universities have to pay for education in Tai-wan, it is
impossible to design personalized learning environments to
accommodate each students’ learning needs. It is possible,
how-ever, by using a decision tree algorithm, for teachers to
investigate students’ learning characteristics in advance, and on
the basis of this information to extract students’ optimal learning
sequences, and
then maximize students’ learning outcome by
grouping students with the same learning sequence together. This
paper will apply a decision tree algorithm, a data mining
technique, to investigate each students’ background and
characteristics in order to optimize his or her learning sequence
and maximize his or her learning outcome in the field of Teaching
English as a Second Language (TESL).
2. Literature review[/B]
2.1. Adaptive learning[/B]
Since conventional classroom learning hinders
students’ poten-tial performance outcomes, it is necessary to
devise techniques of adaptation and personalization in order to
improve their learning process. Gilbert and Han (1999) proposed the
‘‘Case-Based Reason-ing” (CBR) system, according to which new
students, depending on their prior learning experience, would be
assigned to one of four groups – the one deemed most suitable for
providing them with adaptive learning materials and maximizing
their learning out-come. Shang, Shi, and Chen (2001) argued the
necessity of creating an intelligent learning
environment, one that would be student-centered, self-paced, highly
interactive, and based on students’ learning characteristics,
including background knowledge and learning style. Trantafillou,
Poportsis, and Demetriadis (2003) pro-posed an adaptive learning
system called AHS (Adaptive Hyperme-dia System). In the AHS
learning system, students would be divided into two groups with
different cognitive learning styles: one group for student
demonstrating field independence, the other for those showing field
dependence. Each of these proposals takes account of students’
different learning styles, learning characteristics, and learning
preferences in order to help them absorb course material more
quickly and effi-ciently ( Adler & Rae, 2002; Corno & Snow,
1986; Karagiannidis,Sampson, & Cardinali, 2001 ). The advantage
of adaptive learning
is that it offers flexible solutions by
dynamically adapting content to each individual’s learning
needs.
2.2. Learning sequence[/B]
While using adaptation and personalization
techniques to im-prove the learning process for students,
instructors should also consider the sequence in which course
material is taught, for it may lie at the heart of the students’
learning process. According to the IMS (Simple Sequencing
Specification), version 1.0 (2007) definition, sequencing is a
predictable, consistent ordering of course material that delivers
learning activities in an instructional-ly meaningful manner,
without consideration of the delivery envi-ronment. The learning
sequence can be specified by either the course instructors or the
courseware designers. As for the sequence of the instructional
activities, it would be designed after taking full account of the
students’ learning behaviors and backgrounds (Co-lace, De Santo,
& Vento, 2005 ). For learning sequence has an effect on
navigational elements, while teachers choosing the course con-tent
must map a sequence based on students’ characteristics in or-der to
facilitate the learning process. That is, teachers should arrange
different learning sequences to match each individual stu-dents’
portfolio and learning content. Carchiolo, Longheu, and Malgeri
(2002)have proposed utilizing adaptive formative paths in a
Web-based e-learning environment, using domain database and student
profiles to generate a students’ personalized learning path. Taking
into account each students’ prior knowledge and learning
characteristics, the learning se-quences are dynamically modified
to match the students’ needs and capacities. Therefore, the
adaptive learning sequence system is effective in improving
students’ learning achievement.
2.3. Student characteristics[/B]
In order to implement the adaptive learning
sequence in the teaching of English, students’ characteristics or
profiles should be analyzed. It has been shown that the analysis of
students’ learning characteristics and profiles can help teachers
understand the rea-sons why students get high or low grades, by
revealing the implicit rules students follow during the learning
process ( Sarasubm, 1998; Su, Tseng, Wang, & Weng, 2006 ). One
of the most telling factors to be considered is whether a students’
cognitive style is field depen-dence or field independence (Witkin,
1962; Witkin, Moore, Goo-denough, & Cox, 1977). Many
researchers have demonstrated the impact of field dependence/field
independence cognitive styles on students’ learning ( Abraham,
1983; Brumby, 1982; Jamieson & Chapelle, 1987; Summerville,
1999; Witkin, 1962; Witkin et al., 1977). It has also been
demonstrated that motivation plays an important role in learning a
foreign language (Manolopulou-Sergi, 2004; Robinson, 2003; Skehan,
1998 ). That is, providing stu-dents with the proper motivation can
enhance their ability to learn a new language. In addition, the
combination of learning styles and teaching strategies has been
carried out in a variety of educational contexts ( Chen, Liu, Ou,
& Liu, 2000; Chiali, Eberrichi, & Malki,
2006; Evans, 2004). While analyzing students’
learning characteristics, Chen et al.
(2000) applied decision tree and data cube
techniques to investi-gate students’ learning behaviors and observe
learning processes to find out the pedagogical rules on students’
learning perfor-mance. By taking account of students’ learning
profiles, Chen’s decision tree and data cube techniques can offer
students with similar learning characteristics and profiles
appropriately person-alized recommendations. Furthermore, student
profiles can be used in adapting course materials and learning
sequences (Chiali et al., 2006; Evans, 2004; Sarasubm,
1998).
2.4. Data mining[/B]
Data mining is a technique for uncovering hidden
patterns in the object or process
in the data. The uses of data
mining include the following. In the first place, by using the
technique of data mining, a little data can be made to reveal many
new patterns
and new relationships. Second, data mining can
disclose new ways to classify data and can find clusters and
associations within data. Third, data mining can discover new ways
of facilitating better decision making (Devroye, Gyprfi, &
Lugosi, 1997 ). Romero, Ven-tura, and Bra (2004) used evolutionary
algorithms as a data mining
technique to discover interesting relationships in
students’ usage data, which may be very useful to both teachers and
course design-ers in maximizing the effectiveness of a given
course. Lee (2005) proposed a student model in the context of an
integrated learning environment in which diagnostic, predictive,
and compositional modeling were discussed. Both diagnostic and
predictive modeling are applied to issues of credit assignment and
scalability. Compo-sitional modeling of student profiles is used in
the context of an intelligent tutoring system/adaptive hypermedia
learning system
pattern. Hsia, Shie, and Chen (2006) used data
mining techniques to uncover the preferences and predict the future
choices of Continuing Education students at the Extension Education
Center of a university in Taiwan. Hence, by using the data mining
tech-nique, being referred to as database knowledge discovery, an
im-plicit pattern will be elicited from a volume of data ( Klosgen
& Zytkow, 2002; Su et al., 2006 ).
2.5. Decision tree analysis[/B]
A decision tree is a popular technique used for
supervising. A number of papers have demonstrated the successful
application of decision tree models to real-world problems (Luan,
2002; Pli-onis, 2004; Timmermans & Arentze, 2003; Zalik, 2005).
A deci-sion tree is a tool used in the description, classification,
and generalization of data. It can organize descriptions of data
into more compact form. It can also classify data into groups
sharing similar features and characteristics. And it can be used to
gener-alize and predict the value of dependent variables through
map-ping from observations about independent and dependent
variables to make a conclusion about these variables’ target va-lue
(Murthy, 1998).
A decision tree can thus take the form of an
algorithm that uses information to search for prediction rules and
further analyze the results of classification ( Hsia et al., 2006 ).
In research, decision trees have the following advantages. They can be
effectively ap-plied to all types of data structure, discrete,
continuous, or mixed.
In addition, the prediction rules of a decision
tree are easily inter-preted. Finally, it can predict accurately
even in the case of highly non-linear problems (Hautaniemi, Kharait, Iwabu,
Wells, & Lauf-fenburger, 2005 ). In sum, a decision tree can be
an important tool in data mining.
3. Adaptive learning sequence analysis and
discussions[/B]
The purpose of this paper is to propose an
adaptive system anal-ysis, in order to optimize learning sequences
in the TESL. The researchers applied the decision tree technique
to extract automat-ically the optimal learning sequences of
teaching content, as indi-cated by students’ diverse
characteristics, and performances. In
other words, by using a decision tree, the
researchers were able to derive a learning sequence for students that
was individualized and personalized. The learning sequences
obtained were also opti-mal in terms of teaching costs, as
estimated during expert panel discussions. Fig. 1 shows a flow chart of method
employed in our research.
3.1. Data collections and
pretest/posttest[/B]
In order to obtain the results of the adaptive
system analysis for optimal learning sequences, five factors –
gender, personality type, cognitive style, learning style, and the grades
of the previous semester – were selected as students’
characteristics. Fifty fresh-men participated in the experiment.
The participants were stu-dents in the Psychology Department who
had studied English for at least six years after junior high school.
Before the experiment, every student had to fill out a questionnaire
designed to identify his or her characteristics. The coding of the
five student character-istics is shown in Table 1.
Figure
1
英文论文&翻译" TITLE="赵艳平 英文论文&翻译" />
For instance, the code h ME, SF , FI i signifies
that the student has the characteristics of being mildly extroverted,
of having a sense/ feeling learning style, and a field independent
cognitive style. After filling out the questionnaire, all participants
had to take the pretest to determine their initial level of English
reading comprehension ability. After the pretest, they were given
different handouts with different learning sequences to precede their
learning process.The codes
employed in the handouts – which include the main
idea and
details, inference, critical reading, and vocabulary – are
illus-trated inTable 2. The maximum number of learning sequence
arrangements of main idea and details, vocabulary, inference,
and critical reading is twenty-four. Therefore, twenty-four handouts
with twenty-four learning sequences were derived from
the sequence arrangements. Five experts in TESL participated
in panel discus-sions to decide which handouts/learning sequences
would be most suitable for students in Taiwan. After a series
of such discussions, ten feasible learning sequences were decided
upon. Table 3 shows the ten feasible handouts/learning sequences
deemed suitable for students in Taiwan.
All participants had to take the pretest to
realize their initial le-vel of English reading comprehension
ability. The pretest can also serve as the homogenous test. The
purpose of a homogenous test
is to insure
that all participants have the same degree of skill in Eng-lish
reading comprehension. The researchers used the control
chart
method for product quality to test homogeneity
(Stevenson, 2005).
The control chart, based on normal distribution,
includes the cen-ter line (CL), upper center line (UCL), and lower
center line (LCL). The CL is the average gradeX; where X
is the grade of
ith student.
The UCL isX þ 3S ; where S is the
standard deviation. The LCL is X [1]3S . The
pretest results show that the stu-dents are in fact homogenous, as
all the pretest grades are
between UCL
and LCL. Fig. 2 shows the homogeneity of the students’
pretest results.
After the pretest, the students were randomly
given different handouts with different
learning sequences to precede their learn-ing process. Then, after
six weeks of group instruction and person-alized instruction, a
posttest was implemented to obtain students’ learning outcome.
Using the results of the pretest, the posttest,
and the
learning characteristics questionnaire, the students’
profiles were developed, and then the decision tree
technique was used to ‘‘data mine” each students’ profiles in order
to extract the opti-mal learning sequences for that
student.
3.2. Data mining for Adaptive Learning
Sequence[/B]
The researchers used the decision tree technique
to extract the optimal learning sequences based on students’
learning profiles. The decision tree algorithm used in the study
was revised from Roiger and Geatz’s algorithm (2003), which may
be summarized as follows:
STEP 1. Let T be the set of training
instances.
In this paper, the T set included input entries
and output entry. The input entries were the students’ learning
characteristics, including gender, personality type, cognitive
style, learning style, and grade from the previous semester. The output
entry was the optimal learning sequence, which is selected
from posttest minus
pretest >0. Out of the total of 50 students, 41
students fit thisT set.
STEP 2. Choose an attribute that best
differentiates the instances contained in T.
The authors set the best differentiates based on
information gain ratio (IGR) ( Mitchel, 1997). In the input
entries, for example, if the ‘personality type’ characteristic
contributed more to IGR than the other 3 entries, ‘personality type’ would be
selected as the first attribute. Then,In , MI, N, ME, and E would be
separated for the next STEP.
STEP 3. Create a tree node whose value is the
chosen attribute.
Create child links from this node where each link
represents a unique value for the chosen attribute. Use the
child link values to further subdivide the instances into subclasses.
In STEP 2, the 3rd node was personality type, and the child links
wereIn , MI, N, ME, and E. Continuing the child links, the other 3
input entries are con-sidered the IGR contribution for output
entry. Then, the next node for different child links is
obtained.
STEP 4. Repeat each subclass created in STEP
3.
Repeat the node and child link obtained from STEP
3 until all the five characteristics are tested as the
node. With
the application of the decision tree algorithm to the re-sults of
the pretest, the posttest, and the learning
characteristics questionnaire, the optimal learning sequences
based on students’profiles
were compiled and are summarized in Table 4.
From the above table, we can see that for
introverted students can use either C6 or C10. Student with neutral
personalities, and who got lower grades the previous semester, may
choose C1, C3,C8, or C9.
Those students with neutral personalities, but who
got intermediate grades the previous semester, may
choose C1, C2, or C5. Those students with neutral personalities
who got higher grades the previous semester may choose either
C8 or C10. For stu-dents with mildly introverted or mildly
extroverted personalities, the subdivisions of the decision trees would be
more complicated than those of the decision trees of students
with either clearly introverted or neutral
personalities. Mildly introverted students with a
sensing/thinking learning style and a preference for a field independent
cognitive style may choose either C4 or C10. If they have a field
dependent cognitive
Table 1
Learners’ characteristics code
英文论文&翻译" TITLE="赵艳平 英文论文&翻译" />
英文论文&翻译" TITLE="赵艳平 英文论文&翻译" />
Table 2
Handout content
code
英文论文&翻译" TITLE="赵艳平 英文论文&翻译" />
Table 3
The handouts of
learning sequence
英文论文&翻译" TITLE="赵艳平 英文论文&翻译" />
Fig. 2. The homogenous test for
pretest.
英文论文&翻译" TITLE="赵艳平 英文论文&翻译" />
style, they may choose either C6 or C9. Mildly
introverted malestu-dents with a sensing/feeling learning style may
choose C2, C4, or C8. Females with the same characteristics may
choose C1 or C9. Mildly introverted student with an
intuition/feeling learning
style may
choose C2, C3, or C10 learning sequence. Mildly
introverted students with an intuition/thinking learning
style may choose either C2 or C8. Mildly extroverted students
with a sensing/think-ing learning style may choose C2, C3, or C5.
If a student has a sens-ing/feeling learning style and a field
independent cognitive style, he or she may choose C4, C5, or C6. On the other
hand, if he or she is a field dependent learner, the options
are C3, C6, or C10. Mildly extroverted students with an
intuition/feeling learning
style may
choose either C1 or C4 learning sequence. Finally,
mildly extroverted students with an intuition/thinking
learning style may choose either C3 or C8.
Interestingly, it can be seen from Table 4 that
there are no stu-dents who identified themselves as extroverted.
The decision tree analysis also showed that the C7 learning
sequence – critical read-ing ?main idea
?details?vocabulary?inference – does
not occur as
an optimal learning sequence. It is possible that
Taiwanese students are overly shy. In class, they sit
quietly in rows and pas-sively copy down whatever the teacher tells
them ( Rendon, 2005). It appears to be difficult for Taiwanese
students to start a learning sequence with critical reading, since
the latter requires
students to demonstrate their ability to organize,
synthesize, and criticize the material they are reading, and to
express their agree-ment or disagreement with it.
Numerous researchers have shown that critical
reading is one of the most effective ways to solidify students’
prior knowledge; it can also help teach them how to construct an
argument (Case, 2002; Paul & Elder, 2001; Scriven &
Paul, 2003; Taylor &
Patterson, 2000), and give them the opportunity to
articulate their own origi-nal ideas ( Walstad & Becker, 1994).
Yet it is difficult for
Taiwanese students to take advantage of the benefits of
learning to read crit-ically. Being rooted in the hierarchical and
group-oriented culture of China, Taiwanese students have been trained
to listen quietly and attentively to their elders (Hong, Veach,
& Lawrenz, 2004),
and never dare to express their own opinions (
Biggs & Tang, 1996). Being raised in this traditional-bound
and authoritarian cul-ture, students in Taiwan are reluctant to
express their opinions in class, fearing that any talking may at all may
be construed as ‘‘talk-ing back” ( Freire, 1970) to teachers, which
would cause their fam-ilies to lose face. Little by little they
become less adventurous and less outgoing in nature. Thus, it is easy to see
why no students would identify themselves as
extroverts.
In addition, some personalized groups can adopt
more than one learning sequence. For instance, mildly
introverted students with an intuition/feeling learning style may choose
the C2, C3, or C10 optimal learning sequence. In order to save on
teaching costs and facilitate the teaching process, the same five
experts deciding on the feasible handouts/learning sequences
participated in the panel discussions. After these further discussions,
the hierarchical deci-sion tree of the optimal learning
sequences/handouts was simpli-fied, and the optimal learning
sequences/handouts were
minimized to five learning sequences/handouts. The
simplified optimal learning sequences/handouts are as
follows:
1. For
students with neutral personalities
who got lower grades the previous
semester, mildly introverted female
students with a sense/feel in learning style,
and mildly extroverted students with an
intuition/feeling learn-ing style, the optimal
learning sequence handout is C1.
2. For
students with neutral personalities
who got intermediat grades the previous
semester, mildly introverted students with
an intuition/ feeling learn-ing style,
and mildly extroverted students
with a sensing/thinking learning style, the optimal
learning sequence/handout is C2.
3. For
mildly introverted students with a
sensing/thinking learning style and a field dependent cognitive style,
and mildly extroverted students with a
sensing/feeling learning
style and a field independent cognitive
style, the optimal learning
sequence/handout is C6.
4. For
students with neutral personalities
who got higher grades the previous semester,
mildly introverted male students
with a sensing/feeling learning
style, mildly introverted
students with an intuition/thinking learn-ing style,
and mildly extroverted students
with an intuition/thinking learn-ing
style, the optimal learning
sequence/handout is C8.
5. For
introverted students, mildly
introverted students with a sensing/thinking learning
style and a field independent cognitive style,
and mildly extroverted students with a
sensing/feeling learning style and a field dependent cognitive
style, the optimal learning
sequence/handout is C10.
4. Conclusion and future research
The purpose of this paper was to propose an
adaptive system analysis for optimizing learning sequences, in
which a decision tree algorithm, based on student profiles, was used
to extract the most adaptive learning sequences. By applying a
decision tree data min-ing technique to the students’ profiles,
nine optimal learning se-quences for personalized learning were
derived, and the students were grouped into fifteen optimal personalized
learning groups. In order to cut teaching costs and facilitate
the teaching process, after expert panel discussions, the hierarchical
decision tree of
optimal learning sequences/handouts was
simplified, and the sim-plified optimal learning sequences/handouts
were minimized to five learning sequences/handouts.
Table 4
The decision table for learning
sequences Personality
Learning style
英文论文&翻译" TITLE="赵艳平 英文论文&翻译" />
This study has outlined a way in which the process
of learning English can be facilitated for Taiwanese
students who, for cultural reasons, are lacking in both motivation and
self-confidence. The decision tree algorithm technique and theory
discussed in this study could be an important innovation in the field of
TESL in Taiwan. The researchers hope that this paper may serve as an
initial research model for using data mining
techniques to adapt course content
to the
learning needs of individual students. A future study should fo-cus
on designing and developing a way to test the
recommendation rules for optimal learning sequences outlined in
this paper.
References [/B]
Abraham, R.
(1983). Relationship between use of the strategy of monitoring
and cognitive style. Studies in Second Language
Acquisition, 6 , 17–32.
Adler, C., & Rae, S.
(2002). Personalized learning: The future of e-learning is
learner-centric environment. E-learning (January),
22–24.
Biggs, J., & Tang, C.
(1996). Hong Kong students cope with assessment. In D.
Watkins & J. Biggs (Eds.), The Chinese learner:
Cultural, psychological and
contextual influences (pp. 159–182). Victoria, Australia:
Australian Council of
Educational Research.
Brumby, M. (1982).
Consistent differences in cognitive style shown for
qualitative biology problem solving. British Journal of
Educational Psychology, 52 , 244–257.
Carchiolo, V., Longheu, A.,
& Malgeri, M. (2002). Adaptive formative paths in a web-based
learning environment. Educational Technology and Society, 5 ,
64–75.
Case, R. (2002). Partnering
to promote critical thinking. School Libraries in
Canada, 22(1), 11–12.
Chen, G. D., Liu, C. C., Ou,
K. L., & Liu, B. J. (2000). Discovering decision
knowledge from web log portfolio for managing web based
classroom teacher. Journal of Educational Computing Research, 19(3),
307–328.
Chen, J. F., Warden, C. A.,
& Chang, H. T. (2006). Is English a brand: The impact
of English language on product. The Journal of
Language for International
Business, 17(1), 29–42.
Chiali, S., Eberrichi, Z.,
& Malki, M. (2006). A dynamic Composition of an
adaptive course. Transactions on Engineering, Computing
and Technology, VII
(February), 103–108.
Colace, F., De Santo, M.,
& Vento, M. (2005). A personalized learning path
generator based on metadata standards. International
Journal on E-learning, 4 (3), 317–335.
Corno, L., & Snow, E.
(1986). Adapting teaching to individual differences
among learners. In M. Wittrock (Ed.), Handbook of
research on teaching(pp. 605–629).
New York, NY:
Macmillan. Devroye, L., Gyprfi, L., & Lugosi, G. A.
(1997).Lugosi, Probabilistic theory of pattern
recognition . New York, NY:
Springer-Verlag. Evans, C. (2004). Exploring the relationship
between cognitive style and
teaching style. Educational Psychology, 24(4),
509–530.
Freire, P. (1970). Pedagogy
of the oppressed . New York: Herder and
Herder. Gilbert, J. E., & Han, C. Y. (1999).
Adapting instruction in search of ‘‘a
significant difference”. Journal of Network and Computer
Application, 22, 149–160.
Hautaniemi, S., Kharait, S.,
Iwabu, A., Wells, A., & Lauffenburger, D. A.
(2005). Modeling of signal-response cascades using
decision tree analysis. Bioinformatics, 21(9), 2027–2035.
Hong, Z., Veach, P. M.,
& Lawrenz, F. (2004). An investigation of the
gender stereotyped thinking of Taiwanese secondary
school boys and girls. Sex
Roles, 48(11/12), 495–504.
Hsia, T. C., Shie, A. J.,
& Chen, L. C. (2006). Course planning of extension education
to meet market demand by using data mining
techniques: An example of
Chinkuo Technology University in Taiwan. Expert Systems
with Applications. doi:
10.1016/ j.eswa.2006.09.025.
IMS (2007). IMS simple
sequencing specification, IMS Global Learning
Consortium. Accessed
20.06.07.
Jamieson, J., &
Chapelle, C. (1987). Working styles on computers as evidence
of second language learning strategies. Language
Learning, 37 , 523–544.
Karagiannidis, C., Sampson,
D., & Cardinali, F. (2001). Integrating
adaptive educational content into different courses and
curricula. Educational Technology and Society, 4(3).
karagiannidis.html >
Accessed 20.05.07.
Klosgen, W., & Zytkow,
J. (2002).Handbook of data mining and knowledge
discovery.
New York: Oxford University
Press. Lee, C. (2005). Diagnostic, predictive and
compositional modeling with data
mining in
integrated learning environments. Computers and Education. doi:
10.1016/ j.compedu.2005.10.010.
Luan, J. (2002). Data mining
and its applications in higher education. New
Direction for Institutional Research, 13 ,
17–36.
Manolopulou-Sergi, E.
(2004). Motivation within the information processing
model of foreign language learning. System, 32 ,
427–441.
Mitchel, T. M.
(1997).Machine learning. New York: McGraw-Hill.
Murthy, S. K. (1998).
Automatic construction of decision trees from data: A
multi-disciplinary survey.Data Ming and Knowledge Discovery, 2(4),
345–389.
Paul, R., & Elder, L.
(2001).The miniature guide to critical thinking: Concept and tools
. Dillon Beach, CA: The Foundation for Critical
Thinking.
Plionis, E. M. (2004).
Decision tree: A conceptual tool for best practices.
Brief Treatment and Crisis Intervention, 4(1),
37–47.
Rendon, L. (2005). Realizing
a transformed pedagogical dreamfield:
Recasting agreements for teaching and learning.
Spirituality in Higher Education, 2
(1), 1–13.
Robinson, P. (2003).
Attention and memory during SLA. In C. J. Doughty & M. H.
Long
(Eds.), The handbook of
second language acquisition (pp. 631–679).
Oxford:
Blackwell
Publishing.
Roiger, R. J., & Geatz,
M. W. (2003). Data mining: A tutorial-based primer .
Boston:
Wesley.
Romero, C., Ventura, S.,
Delgado, J. A., & Bra, P. D., (2007). Personalized
links
recommendation based on data
mining in adaptive educational hypermedia
systems. In E. Duval, R.
Klamma, & M. Wolpers (Eds.), Creating New
Learning
Experiences on a Global
Scale (Proceedings 2nd European Conference on
Technology
Enhanced Learning, EC-TEL
2007, Crete, Greece, 17-20 September, 2007) Vol.
4753.
Lecture Notes in Computer
Science (pp. 292–306). Crete, Greece: Springer.
Romero, C., Ventura, S.,
& Bra, P. D. (2004). Knowledge discovery with
genetic
programming for providing
feedback to courseware authors. User Modeling
and User – Adapted Interaction, 14,
425–464.
Sarasubm, K. (1998).
Learning style perspectives: Impact in the classroom . Kansas
City:
Atwood
Publishing.
Scriven, N., & Paul, R.
(2003). Defining critical thinking.
criticalthinking.org/page.cfm?PageID=410&CategoryID=51
> Accessed 14.04.06.
Shang, Y., Shi, H. C., &
Chen, S. S. (2001). An intelligent distributed environment
for
active learning.ACM Journal
of Educational Resources in Computing, 1(2),
1–17.
Skehan, P. (1998). A
cognitive approach to language learning. Oxford:
Oxford
University
Press.
Stevenson, W. J.
(2005).Operations management . New York:
McGraw-Hill.
Su, J., Tseng, S., Wang, W.,
& Weng, J. (2006). Learning portfolio analysis and
mining
for SCORM complaint
environment. Educational Technology and Society,
9(1),
262–275.
Summerville, J. (1999). Role
of awareness of cognitive style in
hypermedia. International Journal of Educational Technology,
1 (1). ijet/v1n1/summerville/ > Accessed
24.04.07.
Taylor, R. H., &
Patterson, L. (2000). Using information literacy to promote
critical thinking. Teacher Librarian, 28(2),
9–14.
Timmermans, T., &
Arentze, T. (2003). Measuring the goodness-of-fit of decision-tree
models of discrete and continuous activity-travel choice. Methods
and empirical illustration.Journal of Geographical
Systems, 5 (2), 185–206.
Trantafillou, E., Poportsis,
A., & Demetriadis, S. (2003). The design and the
formative evaluation of an
adaptive educational system based on cognitive. Computer
and Education, 41, 87–103.
Witkin, H. A.
(1962).Psychological differentiation: Studies in development. New
York: Wiley.
Witkin,H. A., Moore, C. A.,
Goodenough, D. E., & Cox, P. W. (1977).
Field-dependent and
field-independent cognitive styles and their educational
implications. Review of Educational Research, 47 (1),
1–64.
Walstad, W. B., &
Becker, W. E. (1994). Achievement differences on
multiple-choice and critical
thinking ability tests in economics. The American Economics
Review, 84(2), 193–196.
Zalik, K. R. (2005).
Learning through data mining. Computers Applications
in Engineering
Education, , 13 (1),
60–65.
在英语教学中的自适应学习顺序的数据挖掘[/B]
摘要[/B]：
本文的目的是提出一种自适应系统优化学习顺序分析。分析采用决策树算法，根据学生的个人资料，为特定的教学内心，开发最适应的学习顺序。个人资料建立在前测和后测，以及一些列的学生特点：性别，性格类型，认知风格，学习风格和学生上学期的成绩的基础上。本文针对在传统的英语教学方法中，坚持一个固定的学习顺序的问题，并建议一个规则来设置一个最佳的学习顺序，从而促进学生的学习过程和最大化学习效果。通过使用本文提出的技术，教师可以同时降低他们的教学成本，并且为学生达到最优化的自适应顺序。结果显示，自适应学习顺序的优势在于它将学生自身的特点以及成绩考虑在内，因为这个原因，它构成了在英语作为第二语言的教学领域的一个创新点。
1[/B]、前言[/B]
[/B] 为了在二十一世纪经济全球化下竞争和生存，学生掌握英文交流能力是非常重要的。( Chen, Warden, &
Chang, 2006 ).英语教学的目的-包括包括理解，听，说，读，写能力，是促进学生的未来学术和职业生涯。学生的学习依赖于教室里发生了什么，并且，学生的学习取决于在课堂上，会发生什么情况，在不同的教室可能是不同的认知和学习风格。但是，在台湾以往的学习系统中，英语老师向学生教授相同的内容，不考虑学生的性别，性格类型，认知风格，学习风格以及以往的知识。也就是说，目前的课程在“静态”的学习材料的基础上，而不是“动态”学习材料(Romero, Ventura,
Delgado, & Bra, 2007)。在这种类型的学习系统中，如果学生希望最大限度地提高他们的学习成绩，他们必须调整他们自己来适应课程内容，然而课程内容永远不会调整来适应他们的个人需求和喜好。
课堂上适应学生需求的自适应包括两个重要问题：如何根据每个学生的特点和能力量身定制课程，如何为不同的学生创建，代表，保持拥有合适的相关顺序定义的活动树。不幸的是，因为在台湾，高校需要在教育上投入大量的花费，所以，不可能设计个性化的学习环境以事业每个学生的学习需求。但是，通过使用决策树算法，教师通过调查提前了解学生的学习特点是可能的，并且，基于这些信息提取学生的最优学习顺序，接下来，通过将拥有相似学习顺序的学生组合在一起以使学习成绩最优化。本文将采用决策树算法，数据挖掘技术，调查每个学生的背景和特性，以便在英语作为第二语言教学的领域，
优化他或她的学习顺序，最大限度地发挥他或她的学习成果。
2[/B]、文献综述[/B]
2.1.[/B]自适应学习[/B]
由于传统的课堂学习妨碍了学生的潜在表现的结果，因此，设计自适应和个性化的技术来该是他们的学习过程是有必要的。吉尔伯特和汉（1999）提出了“基于案例的推理（CBR）系统，根据这个系统，新学生将根据他们先前的学习经验，被分配到四个组中的一个，一个被认为最合适为他们提供自适应学习材料，并最大限度地提高他们的学习结果的组。商，石和陈(2001)
认为有必要建立一个智能的学习环境，一个以学生为中心，自定进度，高度互动，并根据学生的学习特点，包括背景知识和学习的特点，学习风格的智能环境。Trantafillou，Poportsis，Demetriadis（2003）提出了一个自适应学习系统叫做AHS（自适应超媒体系统）。在AHS系统中，学生将根据不同的认知学习风格被分为两组，一组学生表现为场独立性，另一组表现为场依存性。
这些建议考虑学生的不同学习风格，学习特点，学习偏好，以帮助他们更快，更好的吸收课程教材( Adler &
Rae, 2002; Corno & Snow, 1986; Karagiannidis,Sampson, &
Cardinali, 2001 )。自适应学习的优势是，它提供了灵活的解决方案，通过动态调整内容每个人的学习需要。
2.2[/B]．学习顺序[/B]
在使用自适应和个性化技术以改善学生的学习过程中，教师也应该考虑教授课程材料的顺序，因为他在学生的学习过程中起到至关重要的作用。根据IMS（简单排序规范）1.0版本（2007）定义，根据IMS（简单排序规范），1.0版（2007年）定义，测序是可预见的，与在有意义的教学方式中提供学习活动的学习材料顺序一致，不考虑生产环境。学习顺序可以被课程导师或者课件设计者指定。至于教学活动的顺序，需要在全面考虑学生的学习行为和背景下进行设计(Co-lace, De Santo, &
Vento, 2005 )。由于学习顺序对导航元素有影响，因此，当教师靴子课程内容是，必须在学生特点的基础上绘制顺序一遍促进学习过程。也就是说，教师应安排不同的学习顺序，以符合每个学生的组合和学习内容。
Carchiolo，Longheu，Malgeri（2002）提出了在一个基于Web的电子学习环境，利用自适应形成的路径，使用域名数据库和学生资料，以产生一个学生个性化的学习路径。考虑到每一个学生先验知识和学习特点，学习顺序进行动态修改，以符合学生的需要和能力。因此，自适应学习顺序系统有效地提高学生的学习成绩。
2.3.[/B]学生特点[/B]
为了实现英语教学中的自适应学习顺序，应该分析学生的特点和资料。结果显示，对学生的特点和资料进行分析，通过揭示学生在学习过程中隐含的规则，能帮助老师理解学生得高分或低分的原因( Sarasubm, 1998;Su,
Tseng, Wang, & Weng, 2006 )。其中一个最有说服力的因素被认为是一个学生的认知风格是否是场依存型或者场独立型(Witkin, 1962; Witkin,
Moore, Goo-denough, & Cox, 1977)。许多研究人员已经证实的场依存/场独立认知风格对学生的学习的影响( Abraham, 1983; Brumby,
1982; Jamieson& Chapelle, 1987; Summerville, 1999; Witkin,
1962; Witkinet al., 1977)。同时也证实了，动机在学习外语过程中发挥了重要作用(Manolopulou-Sergi, 2004;
Robinson, 2003; Skehan, 1998 )。也就是说，为学生提供正确的动机能强化学生学习一门新语言的能力。另外，在各种教育环境下，学习方式和教学策略，相结合的方式已经进行了( Chen, Liu, Ou, &
Liu, 2000; Chiali, Eberrichi, & Malki,2006; Evans,
2004)。
在分析学生的学习特点时，陈等人（2000）运用决策树和多维数据集技术来研究学生的学习行为和观察学习过程，以便找出学生的学习表现的教学规则。将学生的学习概况考虑在内，陈等人的决策树和多维数据集技术，可以为具有相似学习特点和学习资料的学生提供适合的个性化建议。另外，学生资料可以用于自适应课程资料和学习顺序中(Chialiet al., 2006;
Evans, 2004; Sarasubm, 1998)。
2.4.[/B]数据挖掘[/B]
数据挖掘是一种揭示数据中描述的对象或过程中的隐藏模式的技术。[/B]使用数据挖掘包括以下内容。第一，通过使用数据挖掘技术，很少的数据就可以揭示了许多新的模式和新的关系。第二，数据挖掘可以透露的新方法对数据进行分类，可以发现聚类和数据内部的联系。第三，数据挖掘可以发现新的方式，促进更好决策(Devroye, Gyprfi, &
Lugosi, 1997 )。
Romero, Ven-tura, Bra
(2004)用进化算法作为数据挖掘发现在学生运用情况数据中的有趣的关系，这可能对于教师和课程设计者在一个给定的课程的有效性最大化方面是非常有用的。Lee
(2005)提出了一个在整合学习环境背景下的学生模型，在这个模型中讨论了诊断、预测以及构图造型。诊断和预测建模都应用到了信用分配和可扩展性问题中。学生资料的构图造型在智能教学系统/自适应超媒体学习系统模式的背景下使用。Hsia, Shie, and Chen
(2006)使用数据挖掘技术来发现在台湾一所大学的扩展教育中心的继续教育学生的喜好和预测未来的选择。因此，通过使用数据挖掘技术，又被称为数据库知识发现，将会从一系列数据中发现隐藏的模式( Klosgen
&
Zytkow, 2002; Su et al., 2006 )。
2.5.[/B]决策树分析[/B]
决策树是一种用于监督的流行的技术，大量论文已经证明了决策树模型对现实世界的问题的成功运用(Luan, 2002;
Pli-onis, 2004; Timmermans & Arentze, 2003; Zalik,
2005)。决策树是一种用于描述，分类，概括数据的工具。[/B]它可以将数据描述组织成更紧凑的形式。它还可以将拥有相似功能和特性的数据进行分类成组。并且，它可以用来通过观察独立和从属变量的映射来概况和预测因变量的值，从而对变量的目标值做出决定(Murthy,
1998)。因此，决策树可以采用算法的形式，使用信息来搜索预测规则和进一步分析分类的结果( Hsia et al., 2006
)。在研究中，决策树有如下特点：它们可以有效地适用于所有类型的数据结构，离散的，连续的，或混合的；另外，决策树的预测规则，很容易解释。最后，它可以准确的预测，即使在高度非线形问题中(Hautaniemi, Kharait,
Iwabu, Wells, & Lauf-fenburger, 2005 )。总之，决策树是数据挖掘的一项重要的工具。
3. [/B]自适应的学习顺序分析和讨论[/B]
本文的目的是提出一种自适应系统的分析，以优化TESL中的学习顺序。“研究人员采用了决策树技术自动提取学习内容的最佳学习顺序，由学生的不同特点和表演来指示。换言之，通过使用决策树，研究人员能够推导出学生个体化和个性化的学习顺序。学习顺序的获得也使教学成本最优化，就像专家小组在讨论中估计的一样。图1列出了在我们的研究中采用的方法。
3.1. [/B]数据收集和前测[/B]/[/B]后测[/B]
[/B]
为了得到最优学习顺序的自适应系统分析的结果，五个因素-性别，性格类型，认知风格，学习风格，和以前的成绩学期
-
被选定为学生的特点。50个新生参与了这项实验。参与者是心理学系的学生，他们从初中毕业以后至少学习了六年英语。实验之前，每个学生需要填写一份用来确定他或她的特征的问卷。5项学生特征的编码体现着表1中。
例如，编码表示学生有轻度外向的特点，感知的学习风格，并且具有场独立性的认知风格。在填写完问卷之后，所有的参与者需要进行前测来确定他们原有的英语阅读理解能力。前测之后，他们分别给予不同的讲义与不同的学习顺序，先于他们的学习过程。讲义中采用的编码-包括主要思想和细节，推理，批判性阅读和词汇-在表2中进行解释。
主要思想细节，推和理，批判性阅读和词汇学习顺序安排的最大数目为(1*2*3*4=24)。因此，从顺序安排中得到有24种学习顺序的24种讲义。在TESL中的五位专家参加了决定适合台湾学生的讲义/学习顺序的小组讨论。经过一系列的讨论，确定了10个简单的学习顺序。表3示出了十种可行认为适合台湾学生的学习讲义/顺序。
所有参与者都采取预试，以认识到他们的英语阅读理解能力的初始水平。前测也可以用来作为同质测试，同质的测试的目的是以确保所有与会者都有相同程度的英语阅读理解的技能。研究人员使用为产品质量开发的控制图方法来测试同质化(Stevenson,
2005)。控制图，基于正态分布，包括中心线（CL），上中心线（UCL），下中心线（LCL）。CL是平均
，在这里，
表示第i个学生的分数。UCL为
，在这里，
，是标准差。LCL是
。前测的结果显示，所有的学生都确实是同质的，因为所有前测的分数都在UCL和LCL之间。图2显示了学生前测结果的同质性。
前测之后，每个学生被随机的分配拥有不同学习顺序的讲义来优化学习过程。接下来，经过六周的小组教学和个别化教学，实施后测来得到学生的学习结果。使用前测结果，后测结果，学习特征问卷，学生资料被发展起来，接下来，使用决策树技术对每个学生的资料进行数据挖掘，以便为那个学生提取最优的学习顺序。
3.2. [/B]自适应学习顺序的数据挖掘[/B]
研究人员采用决策树技术基于学生的资料来提取最优学习顺序。本文所使用的决策树算法是对Roiger and Geatz’s
algorithm(2003)进行了修改，它被总结如下：
第一步.让T作为训练实例的设置。
在本研究中，T包括输入项和输出项。输入项是学生的学习特征，包括性别、性格类型，认知风格，学习风格和上学期的分数。输出项是最优学习顺序，是从后测-前测大于0中选出来的。50是学生中的41个学生满足T设置。
第二步.选择一个最好与T设置中情况下不同的属性。
作者基于信息获得比设定最佳的区分(IGR) ( Mitchel, 1997)。在输入项中，例如，如果“性格类型的特点对IGR的贡献比其他3项多，“性格类型”将作为第一选择属性。接下来，In , MI, N, ME,
和 E将在下一步区分开。
第三步.
创建一个树节点，它的值是所选择的属性。
从这个节点创建子链接，其中每个子链接表示所选择的属性的独特的价值。使用子链接值进一步将实例细分到子类。在第二步中，第三个节点是个性类型，子链接是n , MI, N,
ME,和
E，继续子链接，其他的三个输入项被认为是IGR对输出项的贡献。接下来，得到不同子链接的下一个节点。
第四步.重复第三步创建的每一个子类。
重复第三步得到的节点和子链接，知道所有的五个特性都作为节点被测试。
通过决策树算法对前测结果，后测结果，学习特征问卷的运用，基于学生资料的最优学习顺序被编制，并呈现在表4中。
通过以上的表格，我们可以发现，对于性格内向的学生，可以选择C6或者C10。中性个性，并且在之前的学期得到较低分的学生可以选择C1，C3，C8或者C9。中性个性，并且在之前的学期得到中等分的学生可以选择C1，C2或者C5。中性个性，并且在之前的学期得到较高分的学生可以选择C8或者C10。性格轻度内敛的，或轻度外向的学生的决策树的细分会比那些有明确的内敛或中性个性学生的决策树更复杂。
轻度内敛的学生与感测/思考学习的风格和偏好场独立型认知风格的学生可以选择C4或C10。如果他们具有场依存性认知风格，可以选择C6或者C9。
轻度内敛，学习方式是感知/感觉的男学生可以选择C2，C4或者C8。拥有相似特点的女生可以选择C1或者C9。轻度内敛学习方式为直觉/感觉的学生可以选择C2，C3或C10的学习顺序。轻度内敛学习方式为直觉/思考的学生可以选择C2或C8。轻度外向学习方式为感测/思考的学生可以选择C2，C3或C5。如果一个学生拥有感知/感觉的学习风格和场独立认知风格，他或她可能会选择C4，C5或C6。另一方面，如果他或她是场依存性学习者，选择可能是C3，C6或者C10。轻度外向，学习风格为直觉/感觉的学习者可以选择C1和C4的学习顺序。最后，轻度外向的，学习风格为直觉/思考的学生可以选择C3和C8。
有趣的是，从表4可以看出，没有学生将他们自己定义为外向的。决策树分析同时显示，C7学习顺序-批判性阅读-主要思想-细节-词汇-推断-没有作为一种最优的学习顺序发生。可能是因为台湾学生过度害羞。在课堂上，他们静静地坐在座位上，被动地复制老师告诉他们的任何内容(
Rendon,2005)。对于台湾学生来说，以批判性阅读开始的学习顺序是很困难的，因为后者需要学生展示他们的组织能力，合成，批评他们正在阅读的材料，并且表达他们对于材料的同意和不同意的地方。
许多研究表明，批判性阅读是一种巩固学生已有知识的最有效的方法；它也可以教他们如何建造一个参数(Case,2002; Paul
& Elder, 2001; Scriven & Paul, 2003; Taylor &
Patterson,2000)，并且让他们有机会表达自己的原始想法( Walstad & Becker,
1994)。然而对于台湾的学生来说，利用批判性阅读的好处是很困难的。植根于中国分层的和集体导向的文化，台湾的学生被训练的静静的聆听并且用心向长辈学习(Hong, Veach, &
Lawrenz, 2004)并且从来不敢表达他们自己的观点( Biggs &
Tang,1996)。在这种传统的束缚和专制的文化中长大，台湾的学生都不愿意在课堂上表达自己的意见，担心任何谈话可能会被解释为跟老师“顶嘴”( Freire,
1970)，这将会使他们的家庭蒙羞。渐渐的，他们变得越来越不具有冒险精神，并且在生活中更内向。因此，很容易看出为什么没有学生将自己定义为外向。
另外，一些个性化的组可以选择一个以上的学习顺序。例如，轻度内敛，学习风格为直觉/情感的学生可以选择C2，C3或者C10的最优学习顺序。为了节约教学花费并且强化教学过程，相同的决定可行的学习讲义/学习序列五个专家参加了小组的讨论。经过进一步的讨论，最佳的学习顺序/讲义的分层决策树进行了简化，最佳的学习顺序/讲义最小化到5个学习顺序/讲义。简化的最优学习顺序/讲义如下：
1.
对于
l
中性的个性，在之前学期得到较低分数的学生。
l
轻度内敛，学习风格为感知/感觉的女学生，并且
l
轻度性格外向，学习风格为直觉/感觉的学生
l
最优学习顺序讲义是C1
2.
对于
l
中性的个性，在之前学期得到较低分数的学生。
l
轻度内敛，学习风格为直觉/感觉的学生
l
轻度性格外向，学习风格为感知/思考的学生
l
最优学习顺序讲义是C2
3.
对于
l
轻度内敛，学习风格为感知/思考，具有场依存性认知风格
l
轻度性格外向，学习风格为感知/感觉，具有场独立性认知风格
l
最优学习顺序讲义是C6
4.
对于
l
中性的个性，在之前学期得到较高分数的学生。
l
轻度内敛，学习风格为感知/感觉的男生
l
轻度内敛，学习风格为直觉/思维的学生
l
轻度性格外向，学习风格为直觉/思维的学生
l
最优学习顺序讲义是C8
5.
对于
l
内敛学生
l
轻度内敛，学习风格为感知/感觉，具有场独立性认知风格的学生
l
轻度性格外向，学习风格为感知/感觉，具有场依存性的认知风格的学生
l
最优学习顺序讲义是C10
4.[/B]结论和未来研究[/B]
[/B]
本文的目的是提出一种自适应系统分析，用来优化学习序列。在这种自适应系统中，人们用基于学生资料的决策树法则提取出最适合的学习序列。运用决策树数据分析技术分析学生的资料，产生了9个理想的个性化学习序列，并把学生分成了15个最佳的个化学习小组。
为了减少教育成本、促进教育发展，经专家小组讨论，简化了分层决策树最佳的学习序列/讲义，并将之精简为5个。
这项研究提出了一个方法，这个方法能强化因为文化原因同时缺少动机和自信的台湾学生的英语学习过程。在这项研究中讨论的决策树算法技术和理论在台湾TESL领域是个重要的创新。研究人员希望本研究可以作为用数据挖掘技术使课程内容适应学生个体学习需要的原始研究模型。未来的研究应该着眼于设计和开发一种测试本文提出的最优学习顺序建议的方法上。
[/B]
|
