International Journal of Recent Technology and Engineering (IJRTE)

ISSN: 2277-3878, Volume-8, Issue-1S4, June 2019

Personalized Learning Model Using Item

Response Theory
Amy Lyn M. Maddalora

Abstract— Many Higher Education Institutions in the will be engaged in e-learning by 2015 and that e-learning
Philippines now accept that a blended approach offers countless does not only save money, it also saves time. Therefore, due
advantages in most areas of learning. However, learner’s ability to the growth of the Internet, an increasing number of
has been neglected as a significant factor in student’s success.
individuals are moving to online learning.
Thus various techniques have been developed such as
personalization to improve the learning process and Not all learners have the same personal preferences and
accommodate diversity of learners. This study introduces a learning styles [2]. Thus, many researchers have explored
personalized learning model that recommends Shortest Learning different techniques, such as shortest learning path to
Sequence (SLS) to remediate students with learning difficulty. improve learning process, as well as to accommodate a
The learning model was created using Item Response Theory diversity of learners. An article written by [3] mentioned
(IRT) implemented in an e-learning environment. Assessments
that the individualized interactivity has contributed
were given during the learning process with levels of difficulty.
IRT probabilistically estimates the student’s proficiency of the significantly to the effectiveness of the e-learning
topics taking into considerations the difficulty of the test items. environment. Generally, most personalized systems have
SLS consists of lessons recommended to students, which were been developed to consider learner preferences, interests,
ranked accordingly. The one parameter (1PL) model was used to and browsing behaviors in providing personalized services.
evaluate the test score and the item information was used to rank However, learner ability [4] and necessary corrective
the lessons. Lessons are reduced until the proficiency level of the
measures to remediate learning difficulties [5] are usually
student is reached. Results show that the personalized learning
model is capable of recommending shortest learning sequences. neglected as important factors in implementing personalized
Hence, reduces the time spent of study. The student’s proficiency mechanisms. As a result, this paper explores a learning
level was increased through the implementation of the technique to remediate learning difficulties of students in an
personalized learning model. Thus, the learning outcome was e-learning environment.
also improved. Course contents have been designed to accommodate
Index Terms—shortest learning sequence; item response teaching approaches that support the learning needs of
theory; e-learning; personalized learning model.
students. This includes students experiencing difficulties in
learning. Students having difficulty learning are classfied
I. INTRODUCTION
into the heterogenous group, as they have a wide variety of
E-learning has long been in existence since 1999. E- characteristics, ranging from academic difficulties to
learning is often considered as a means of permitting access cognitive and socio-emotional problems [6]. The term
to learning by using electronic media such as computers, “learning difficulties” is a general one used widely and
tablets, or mobile phones. Long before the internet was without much precision. Usually, the term applies to
launched, distance courses were being offered to provide approximately 10 to 16% of the school population and refers
students with education on particular subjects or skills. With to students who have general problems in learning [7]. Most
the introduction of the computer and internet in the late 20th students experiencing these difficulties could participate
century, e-learning tools and delivery methods expanded. fully in learning experiences and assessment activities
Since then, several schools had been set up to deliver provided by the proposed learning environment.
courses online and bring education to people who wouldn’t Assessment plays an important role in learning. It
be able to attend college due to geographical or time determines whether the abilities and skills of the students in
constraints. In general, e-learning facilitates the delivery of a particular course have been developed. More specifically,
lectures, the administering of quizzes and examinations, and assessment is the way the instructors gather data about their
the designation of paper works. teaching and their student’s learning. A summative
A growing number of business organizations began using assessment is administered using the e-learning prototype to
e-learning to train their employees. New and experienced determine the learning path of the student. Summative
workers alike now had the opportunity to improve their examination takes place after the learning has been
industry knowledge base and expand their skill sets. Many completed that sums up the teaching and learning process.
e-learning authorities in industry and in schools now accept Questions in the exam are constructed using Anderson and
that a blended approach offers countless advantages in most Krathwohl – Bloom’s Revised Cognitive Taxonomy.
areas of learning, and good practice has been developed in a Anderson [8] mentioned in her paper that the taxonomy
hybrid online model [1]. According to the blog posted by the table emphasizes alignment in terms of student learning and
admin of the website CertifyMe.net, entitled the Important provides in-depth examination of alignment. Questions are
eLearning Statistics for 2013, 50% of all college students

Revised Manuscript Received on June 10, 2019.

Amy Lyn M. Maddalora, DIT, Isabela State University, Cabagan,
(aimee_jc@yahoo.com)

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: A11500681S419/19©BEIESP 811 & Sciences Publication
 PERSONALIZED LEARNING MODEL USING ITEM RESPONSE THEORY

generated randomly from the test bank for each student,

which ensures that no student has the same series of
questions. Multiple choice is the most commonly used type
of question for assessment. However, [9] stated that
different types of questions must be included in asssessment
to encourage deep approaches to learning. Thus, this study
employs other question types apart from the standard
multiple choice type of question.
A lot of studies have proposed personalization that
implements various models like Modified Roulette Wheel Fig. 1 Conceptual framework of the personalized
Selection algorithm [10] which recommends new learning learning model
sequences based on students knowledge and performance;
and Item Response Theory (IRT) which recommends Fig. 2 demonstrates the generation of the shortest learning
appropriate learning contents matching the learner’s sequence or SLS using IRT. Data collected from the results
knowledge and ability [11]. IRT is a theory of how people of the exam are used to compute the student’s ability
respond to items. The purpose of this model is to probabilistically with regards to his/her answers to test
probabilistically explain an examinee’s responses to test items. The responses of the students to the test items are
items via a mathematical function based on his/her ability. then converted dichotomously into 1’s (correct answer) and
This paper explores the use of Item Response Theory to 0’s (incorrect answer). These are stored in a matrix where
recommend the shortest learning sequences ideally suited the columns are the test items and the rows represent the
for students with learning difficulties that would lessen their students. The lesson difficulty is determined using the
time of study. formula of the odds of success, or the number of correct
Thus, this paper provides an alternative way to improve answers in each lesson (1). On the other hand, the learning
the learning process by recommending shortest learning ability of the students is estimated based on the odds of the
sequences and to remediate learning difficulties that lessens success or the number of correct answers in the test (2). In
the time of study. Students take the assessment for the this case, the lesson difficulty and the learning ability have
purpose of this personalized learning model. Different been identified. Using the one-parameter logistic model,
question types are employed to encourage different learning also known as the Rasch model (3), the probability that the
approaches. Moreover, questions are aligned according to student learned a particular lesson is computed based on
Anderson and Krathwohl Cognitive Taxonomy. The results his/her ability to answer the test items in each lesson with
of the assessment are analyzed and processed using the respect to the item difficulty. The result of this determines
formula of One-Parameter Logistic Item Response Theory the lessons that the student needs to review in order for
model. him/her to achieve proficiency of the course. Reinforcement
activity is recommended and new learning sequence is
II. CONCEPTUAL FRAMEWORK OF THE STUDY generated based on the formula of item information function
The purpose of the conceptual framework, as presented in (4).
Fig. 1, is to show the components or steps by which the During the learning process, as the iteration increases, the
researcher implements procedures tailored to the learning learning sequence becomes shorter. This is because the
needs of each student. learnt lessons are removed from the learning sequence,
Initially, the student takes a 50-item diagnostic meaning that only those topics that the student failed to
examination to determine his/her knowledge of the course. answer correctly are recommended for a re-study. The
The test questions are generated randomly from the test iteration process continues until the student passed the
bank. Questions are constructed and categorized using summative examination. However, the generation of the
Anderson and Krathwohl Taxonomy to measure the personalized learning would usually stop at level 3 because
cognitive domain of the students. The result of the test according to [12], the mastery level can be achieved at this
determines the first learning sequence (at level 1). The level due to the reinforcement process.
students read and study the lessons from the e-learning
course materials. After finishing these lessons, the student
takes the summative examination. The result of the
examination will be used to determine the shortest learning
sequence using Item Response Theory (IRT).
IRT evaluates the ability of the student, taking into
consideration the difficulty of the lessons and recommends
the shortest learning sequence to him/her. The student can
be said to have learned his/her lesson if he/she answers the
corresponding test items correctly. Therefore, if the student
obtains a passing mark of 85% in the summative Fig. 2 Shortest learning sequence generation
examination, the learning goal is achieved. Otherwise, the
system will recommend a new learning sequence for the
student to read and study again.

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: A11500681S419/19©BEIESP 812 & Sciences Publication
 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-8, Issue-1S4, June 2019

III. RESEARCH METHODOLOGY the test bank with combinations of easy, moderate and
difficult questions. For the purpose of presentation of this
A. Research Design
study, the 50-item examination was used to evaluate the
The research design refers to all the overall strategy that student’s understanding of the course. Since there are ten
the researcher chooses to integrate the different components (10) lessons in the course and each course generates 5-points
of the study in a coherent and logical way, thereby, ensuring test items, the researcher purposively used the 50-item
to effectively address the research problem. It constitutes the examination to have equal distribution of points from each
blueprint for the collection, measurement, and analysis of lesson. Furthermore, questions can be dynamically added to
data [13]. the item bank.
Descriptive research design attempts to describe and
explain conditions of the present by using many subjects and C. Item Response Theory Model
a questionnaire to fully describe a phenomenon [13]. The Student’s ability cannot be judged based on the number of
researcher used the descriptive research design in this study. correct items obtained. Rather, the item difficulty should
The study describes the use of IRT in recommending the also be taken into account [14]. Consequently, students can
shortest learning sequence. It also describes how the have different levels of ability, and items can differ in many
personalized learning model helped improve the learning aspects – most importantly, some are easier and some are
outcome of the student based on the results of the more difficult [15].
examination. Item Response Theory (IRT) Model is a theory of how
people respond to items. It predicts a certain person will
B. Methods and Techniques Used
give a certain response to a certain item. IRT provides a
The course materials for Introduction to Programming model-based linkage between item responses and the latent
using C++ were used as the learning content. The lessons characteristics assessed by a test or scale [16]. IRT models
are sequenced accordingly with the course syllabus, and their corresponding parameter estimation techniques
reviewed and evaluated by academic coordinators, where have a long history of development in the psychometrics
the researcher taught. The learning content has been the literature.
product of the researcher in her 13 years of teaching the The purpose of these models is to probabilistically
course. explain an examinee’s responses to test items via a
Questions are categorized according to Anderson and mathematical function based on his/her ability. The goal of
Krathwohl Taxonomy, which was published in 2001. Fig. 3 IRT is to estimate the learner’s ability with regards to
illustrates the Revised Bloom’s Taxonomy of the Cognitive his/her dichotomous answers to test items. These test items
Domain defining the levels of thinking. The levels build in are scored dichotomously: the correct answer receives a
increasing order of difficulty from basic, rote memorization score of one, and each of the distractors yields to a score of
to higher levels of critical thinking skills. As shown in the zero. Items scored dichotomously are often referred to as
diagram, three changes have been made in the category binary items [17].
level. Comprehension was renamed Understand, Synthesis Each lesson has difficulty level based on the incorrect
changed places with Evaluation and was renamed Create responses on the examination by the learner. Initially, the
[8]. Test questions are created and validated based on the 13 difficulty level is computed by taking the natural log of the
years teaching experience of the researcher. The test items odds of failure (the number of incorrect answers) given by
were already used in the traditional classroom learning.
Multiple choice question is a widely used and highly 𝟏−𝒑
regarded question type. However, other cognitive domains 𝒃𝒊 = 𝒍𝒏 (1)
𝒑
may not be measured. Therefore, different types of
questions are constructed to encourage different approaches Equation (1) is the difficulty logit [18] where b is the
to learning, which includes true or false questions. difficulty of ith lesson and p is the number of correct
responses to test items. Negative logit means the lesson is
easy, 0 means the lesson is moderately difficult, and positive
logit means the lesson is hard.
Student’s ability is estimated based on the correct
responses on the examination. Ability  is computed by
taking the natural log of the odds of success (the number of
correct answers) given by
𝒑
𝜽𝒋 = 𝒍𝒏 (2)
Fig. 3 Anderson and Krathwol cognitive taxonomy 𝟏−𝒑
The Ability Logit [18] in (2) is the  ability of j student
th

It is a repository where all the questions are drawn for the and p is the number of correct responses to test items.
summative examination. Each question in the item bank Negative logit means the ability level is poor, 0 indicates an
contains the following properties: question name, average ability level, and positive logit means ability level is
description, type, category, and points. Questions from the high.
item bank are selected proportionally for each lesson based
on the revised Bloom’s cognitive taxonomy. For each
lesson, a total of 5 (five) points are randomly picked from

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: A11500681S419/19©BEIESP 813 & Sciences Publication
 PERSONALIZED LEARNING MODEL USING ITEM RESPONSE THEORY

In the One-Parameter Logistic (1PL) Model or Rasch beginning a learning activity. The results of this exam are
Model, the probability of a correct response is determined used by the model to generate the initial learning sequence
by lesson’s difficulty and the student’s ability by of the student.
The results of the summative examination designed for
(𝛉 −𝒃 )
𝒆 𝐣 𝒊 the course are extracted and fed to the model to evaluate the
𝑷(𝜽) = (𝛉 −𝒃 ) (3)
𝟏+𝒆 𝐣 𝒊 understanding degree of a learner. Having finished reading
all the lessons, the student must answer the corresponding
P() determines the probability of a correct answer where test items. The student is considered to have understood the
 is the individual ability level of jth student, b as the lesson if he/she answers them correctly. IRT generates the
difficulty parameter of ith lesson and 𝑒 is an exponential shortest learning sequence by estimating student’s ability,
function equal to 2.718281 [15]. 1PL is the simplest IRT ranks the lessons accordingly and recommends the learning
model for a dichotomous item that has only item parameter sequence to him/her.
– the difficulty parameter. Fig. 4 shows the graph of the one-
parameter logistic function using the Item Characteristic IV. RESULTS AND DISCUSSION
Curve (ICC). The proposed personalized learning model was
implemented in an e-learning environment. A section of 40
college students participated in the conduct of this research.
The course contains 10 lessons in the course, Introduction to
Programming Using C++. For the purpose of presentation, a
50-item diagnostic examination was given to students to
assess their knowledge of the course. Each lesson has a test
composed of a total of 5 points which are randomly selected
from the test bank with a combination of easy, moderate and
hard questions.

TABLE I
Fig. 4 Item characteristic curve EXAMINATION RESULTS
Level of Number of
The IRT model illustrates the relationship between the Scores
Proficiency Students
learner’s answer and a test item through the Item High 40 and above 15
Characteristic Curve. The standard mathematical model for Average 30 - 39 14
the ICC is the cumulative form of the logistic function. It Low Below 30 11
defines a family of curves having the general shape of the
ICC shown in Fig. 4. The horizontal axis represents the Table 1 summarizes the results of the diagnostic
ability of student in a limited range and the vertical axis examination after the test was administered to 40 students.
represents the probability that the student with certain ability Given the passing rate of 80% (40 out 50 correct items),
can answer the test correctly [17]. It shows how the
only few passed the examination. Based on the table, 15
interaction of student ability and item difficulty influences
students passed the examination, that is 37.5% of the total
the predicted probability of a correct response to the item
takers. They are considered with high proficiency level of
[15].
the knowledge domain since the threshold value is higher
The Item Information Function (IFF) is related to the than the normal passing rate, that is 75%. On the other hand,
accuracy with which ability is estimated. IIF provides 62.5% (25 students) of the total population have failed the
information about the ability of the student depending on examination. Thus, these students took remedial of the
how closely the difficulty of the item matches the ability of
course to increase their proficiency and remediate learning
the student. The item information function of the 1PL model
difficulties.
is given by
A. Algorithm
𝑰𝒊 𝜽, 𝒃𝒊 = 𝑷𝒊 𝜽, 𝒃𝒊 𝑸𝒊 𝜽, 𝒃𝒊 (4)

The item information is used to rank the lessons in

recommending the shortest learning sequence where
𝐼𝑖 𝜃, 𝑏𝑖 is the item information of ith lesson, 𝑃𝑖 𝜃, 𝑏𝑖 is the
probability of correct responses on ith lesson, and
𝑄𝑖 𝜃, 𝑏𝑖 = 1 – 𝑃𝑖 𝜃, 𝑏𝑖 the probability of incorrect
responses on ith lesson. As the ability becomes either smaller
or greater than the item difficulty, the item information
decreases.
D. Data Collection
Student takes a diagnostic examination to determine
his/her prior knowledge and misconceptions before

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: A11500681S419/19©BEIESP 814 & Sciences Publication
 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-8, Issue-1S4, June 2019

The pseudocode above shows how the equation were used has the highest learning ability and Student 29 has the
to develop the algorithm for generating the shortest leaning lowest learning ability.
sequence. It begins with comparing the test score to the The values in the rows L1 to L10 represents the difficulty
passing score, which is 80. Then, ability logit and lesson level of each lesson based on the student’s ability. Lesson
difficulty logit are computed assigned to variables a and b difficulty is between -4 and +4. That is, lessons with
respectively. A loop continues to iterate until the value of n difficulty level -4 are viewed as the easiest, lessons with
is less than the values of variable i, the total number of items level 0 are viewed as having moderately difficult, and those
in each lesson. Probability of getting the correct and that are with difficulty level 4 are the hardest. The results of
incorrect responses to test items are estimated, which are Table 2 shows that Lesson 5 is the easiest and Lesson 8 is
stored in the arrays. The item information values are stored the hardest.
in the array info and are arranged in increasing order. These Therefore, the proposed personalized learning model
values are the generated shortest learning sequence considers both lesson difficulty and the learner ability
recommended to students. because these parameters affect the student learning
outcome.
B. Lesson Difficulty and Student Ability
The results of the examination were used to assess the C. Estimation of Student Learning Abilities
learning ability of the student in each lesson. The one- The probability that the learners can completely
parameter logistic (1PL) IRT model, also known as Rasch understand (P()) and cannot understand (Q()) the lesson
model, was used to recommend the shortest learning at a level with respect to their ability level are estimated at
sequence based on their individual ability taking into this point. The acceptable threshold value is 0.80
account the difficulty of the lessons. proportionate to the examination passing rate.

TABLE 2 TABLE 3
LESSON DIFFICULTY AND STUDENT ABILITY PROBABILITY OF CORRECT RESPONSES

Table 2 shows the lesson difficulty and the student ability

based on the correct and incorrect responses of the student’s After determining the lesson difficulty and the estimated
answer to the test items. First, the proportion incorrect is ability of the student in the course, the probability that the
calculated by getting the average of student’s incorrect student can respond correctly to the test items in each lesson
responses for each lesson. The proportion incorrect is then given the passing rate of 0.80 is identified. This is computed
converted into lesson difficulty logits by taking the natural using the values derived from (1) and (2), and the formula in
log of the odds of failure using (1). For example, the (3).
difficulty level of each lesson for Student 4 are L1=1.39, As shown in Table 3, scores 0.80 and above means that
L2=-0.41, L3=-1.39, L4=-4, L5=-0.41, L6=0.41, L7=1.39, the proficiency level of the lesson has been achieved by the
L8=4, L9=-0.41, L10=-0.41 respectively. student. Otherwise, if the computed score is below the
On the other hand, the student ability is estimated by threshold value, the student reviews the lesson again with a
calculating the average proportion correct in the new learning sequence recommended by the model. The
examination. Converting the proportion correct into student items highlighted in red are the lessons that need to be
ability logits is done by taking the natural logs of the odds of reviewed by the student. These values are the Item
success using (2). As shown in Table 2, Student 4 has the Information derived from the probability that the students
ability logit of 0, which means that the student has an can completely understand (P()) and cannot understand
average ability level. (Q()), the lesson with respect to their ability level using the
The Ability column in Table 2 shows the student’s ability (4). The item information is sorted in ascending order that
level of the course. Ability greater than 0 indicates that the
student has high proficiency level, ability 0 means an
average proficiency level, and ability less than 0 indicates
that the student has low proficiency level. Hence, Student 32

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: A11500681S419/19©BEIESP 815 & Sciences Publication
 PERSONALIZED LEARNING MODEL USING ITEM RESPONSE THEORY

determines the new learning sequence recommended to the

learners. For example in Table 3, the new learning sequence
for Student 4 is L1L2L3L4L5L6L7L8 L-
9L10. Therefore, those lessons with scores 0.80 and above
are eliminated from the learning sequence, which makes the
study time shorter.
Summative examination is given at the end of the course
review. As the learning ability increases, the more likely the
student can understand the lesson. Therefore, the proficiency
level also increases.
Based on Table 3, Student 32 with the highest ability
level among the 25 students who failed the examination, has
passed all the lessons except for Lesson 8 since it shows that
this Lesson is the hardest of all. And Student 29, which was Fig. 6 Item Information Curve
identified earlier with the lowest ability level, has to review
all the lessons in the course. TABLE 4
SHORTEST LEARNING SEQUENCES

Fig. 5 Relationship between the lesson and student

ability

The graph in Fig. 5 shows the relationship between the

difficulty level of the items in each lesson and the learning
Based from the results of the responses of the students to
ability of the student to answer these items correctly or
the tests analyzed by the personalized learning model, the
incorrectly. The higher the ability level, the most likely that
shortest learning sequences were generated. Shortest
he/she can answer the items in the test correctly. Hence, it is
more likely that the student can completely understand the Learning Sequence (SLS) consists of a short series of
lessons recommended for the students who undergo
lesson.
reinforcement process in order to remediate their learning
D. Recommendation of Shortest Learning Sequence difficulty.
The proposed personalized learning model uses the As shown in Table 5, 25 out 40 students underwent
responses of the student for the estimation of his/her ability. reinforcement process. At Level 0, the students read all the
If the student gives correct answers to the test items in each course materials of the same sequence of lessons L 1L2
lesson, then his/her ability increases. Otherwise, the L3L4L5L6L7L8 L9L10. Students then took the
student’s ability will be decreased. Having estimated the diagnostic examination after they read all the lessons.
student’s ability, lessons are sequenced according to the Students who failed the examination proceed to the next
information value illustrated in Fig. 6. This value depends level (Level 1) of reinforcement and the system
on the matching degree between the difficulty of the lesson recommends a shortest learning sequence. Student 4, as
and the student’s learning ability. The information value is presented in Table 5, was given a new shortest learning
used to arrange the sequences of lessons recommended by sequence of L8L1L7L2L5L6L9L10 which is
the system to the student. 80% of the lesson was recommended and 20% was
successfully passed after the reinforcement. The process
continues until the student has successfully passed the
examination, which was achieved at Level 3.

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: A11500681S419/19©BEIESP 816 & Sciences Publication
 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-8, Issue-1S4, June 2019

remediating student learning difficulties, and be able to

address their learning needs by recommending
individualized course sequencing or personalized lessons.
Diagnostic and summative examiantion are taken by the
students to assess their learning of the course. Different
question types are employed to encourage different learning
approaches. Moreover, questions are aligned according to
Fig. 7 E-learning prototype Anderson and Krathwohl Cognitive Taxonomy. The results
of the assessment are analyzed and processed using the
With the results presented, notice that the reinforcement formula of One-Parameter Logistic Item Response Theory
process stopped at Level 3 as shown in Fig. 7. The proposed model. IRT evaluates the ability of the student taking into
personalized learning model guaranteed that the student considerations the difficulty of the lessons and then
would pass the course as the lesson decreases while recommends the shortest learning sequence to him/her. The
increasing the proficiency level of the student. student is considered to learn a lesson if he/she answers the
corresponding test items correctly. Therefore, if the student
TABLE 5 obtains a passing mark of 85% in the summative
OVERALL MEASUREMENT OF EFFECTIVITY OF THE examination the learning goal is achieved. Otherwise,
PERSONALIZED LEARNING MODEL remedial of the course is recommended and new learning
Grade sequence is generated based on the algorithm developed.
Level 0 Level 1 Level 2 Level 3
Remarks B. Conclusion
Passed 15 37.5% 7 28% 16 88.9% 2 100% Researches in the field of personalization have greatly
Failed 25 62.5% 18 72% 2 11.1% 0 0% contributed to the improvement of student learning outcome.
Total 40 25 18 2 Studies have shown that a personalized learning helped
learners to learn more effectively and efficiently.
Results show that the integration of the personalized Using the One-parameter Logistic IRT model (Rasch
learning model to the prototype is considered effective. Model), the proposed personalized learning model can
Among the 40 students assessed in the conduct of this study, estimate the learning ability of the student, taking into
as shown in Table 5, 15 or 37.5% passed the course and did account the difficulty level of the lesson. These two
not undergo remedial process. Nonetheless, 25 or 62.5% of parameters are important factors in estimating the likelihood
the students failed the course and took remedial of the of the student to pass the course. The personalized learning
course at Level 0. model has been successfully implemented in an e-learning
At Level 1, 25 students who failed the examination took environment that recommends the shortest learning
the remedial and 7 or 28% of these takers passed the sequence to achieve learning goals. Based on the related
summative examination which means that they were able to literature, studies have shown that the use of IRT is proven
achieve the proficiency level of the course. While 16 or to improve the learning process to remediate learning
88.9% of the students successfully passed the examination at difficulties.
Level 2. On the other hand, only 2 or 11.1% of the students The model recommends the shortest learning sequence to
failed the summative examination. Lastly, the remaining 2 students who underwent reinforcement process to remediate
takers were able to completely understand the lessons by learning difficulty. The study proved that as the
passing the course at Level 3. This means that the recommended learning sequence decreases, the student’s
personalized learning model is effective in remediating learning ability is increased. Hence, the student is
learning difficulty because of the shortest recommended guaranteed to pass the course after the reinforcement
learning sequence of lessons. process. Previous studies have also shown that based on the
The results of the study have been proven effective in learner’s ability, a personalized learning path can accelerate
improving the learning outcome of the student to remediate learner learning efficiency and effectiveness.
learning difficulties through the implementation of the In general, the proposed personalized learning model
personalized learning model in the e-learning environment. guaranteed that the student would pass the course as the
lesson decreases while increasing the proficiency level of
V. SUMMARY, CONCLUSION AND the student. The results of the study have proven effective in
RECOMMENDATION improving the learning outcome of the student to remediate
A. Summary learning difficulties through the implementation of the
personalized learning model in the elearning environment.
Various techniques have been developed by researchers to
revolutionize the teaching operation and improve learning C. Recommendation
process through personalized learning path. This paper The researcher has drawn the recommendations for future
provides an alternative way to improve the learning process researchers who will conduct related studies, as well as for
by recommending shortest learning sequences and to the enhancement of the proposed personalized learning
remediate learning difficulties that lessens the time of study. model.
This study is of great value to academic institutions to
enhance the teaching and learning method. The proposed
personalized learning model can be used as a means of

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: A11500681S419/19©BEIESP 817 & Sciences Publication
 PERSONALIZED LEARNING MODEL USING ITEM RESPONSE THEORY

The proposed study can be integrated to the e-learning 15. Partchev, I. (2004). A visual guide to item response theory.
systems of educational institutions who wanted to leverage 16. Drasgow, F., & Hulin, C. L. (1990). Item Response Theory.
the use of their system as well as to improve the learning 17. Baker, F. B. (2001). The Basics of Item Response Theory
(Second ed.). USA: ERIC Clearinghouse on Assessmet and
process. This leaning model will be useful to students in Evaluation.
remediating learning difficulty. 18. Moulton, M. H. (2003). Rasch Estimation Demonstration
Further studies are required to improve the personalized Spreadsheet. https://www.rasch.org/moulton.htm
learning model. Moreover, future researchers are welcome 19. Admin. (2013, March 5). Important eLearning Statistics for
to explore other IRT models such as the 2PL and 3PL 2013. Retrieved from
model. Question of different types can be added to CertifyMe.net:http://www.certifyme.net/osha-blog/elearning-
statistics-
encourage different approaches of learning. 2013/#sthash.M0w9GTVM.dpufhttp://www.certifyme.net/osh
The proposed e-learning prototype should be tested and a-blog/elearning-statistics-2013/
implemented to a wide-array of users since it was only
tested on 40 college students. Exploring the possibility of
implementing the personalized learning model as a plug-in
to an existing e-learning environment should also be
considered by future researchers.

REFERENCES
1. Martyn, M. (2003). The Hybrid Online Model: Good Practice.
Educause Quarterly, pp. 18-23.
2. Pappas, C. (2015, November 30). 7 Tips to create personal
learning paths in e-learning. Retrieved March 15, 2016, from
eLearning Industry: http://elearningindustry.com/7-tips-
create-personal-learning-paths-elearning
3. Capper, J. (2001, May/June). E-Learning Growth and
Promise. TechKnowLogia, 7-10.
4. Chen, C.-M., Lee, H.-M., & Chen, Y.-H. (2005). Personalized
Elearning System Using Item Response Theory. Computers &
Education, 44, 237-255.
5. Ballera, M., Lukandu, I., & Radwan, A. (2015). Reversed
Roullette Wheel Selection Algorithm (RWSA) and
Reinforcement Learning (RL) for Personalizing and
Improving E-Learning System: The Case Study and Its
Implementation. The International Journal of E-Learning and
Educational Technologies in the Digital Media (IJEETDM), 1
(2), 92-108.
6. Kraayenoord, C. V., & Elkins, J. (1990). Learning Difficulties
in A. Ashman and J. Elkins (eds) Education Children with
Special Needs. New York: Prenticec Hall.
7. Louden, W., Chan, L. S., Elkins, J., Greaves, D., & House, H.
(2000). Department of Education, Training and Youth Affairs,
Commonwealth of Australia. Mapping the Territory - Primary
Students with Learning Difficulties: Literacy and Numeracy.
8. Anderson, L. W. and Krathwohl, D. R., et al (Eds.) (2001) A
Taxonomy for Learning, Teaching, and Assessing: A
Revision of Bloom’s Taxonomy of Educational Objectives.
Allyn& Bacon. Boston, MA (Pearson Education Group)
9. Hubbard, R. (n. d.). Questions for Formative and Summative
Assessment that Encourage Deep Rather than Surface
Approaches to Learning Basic Statistics in a Computer
Environment, Queensland University of Technology.
https://www.causeweb.org/cause/archive/artist/articles/hubbar
d.pdf
10. Ballera, M. A. (2015). Adaptive and Personalizing Learning
Sequence Using Modified Roulette Wheel Selection
Algorithm. International Scholarly and Scientific Research &
Innovation, 9 (12), 2265-2269.
11. Kardan, A., & Hosseini, R. (2011). Personalized Content
Sequencing Based on Choquet Fuzzy Integral and Item
Response Theory. International Conference on Education
Technology and Computer, 134-138.
12. Ballera, M. A. (2016). Roulette Wheel Selection Algorithm
and Reinforcement Learning, Lambert Academic Publishing.
13. USC Libraries (2016, April 16). Research Guides, University
of Southern
Carolina.http://libguides.usc.edu/writingguide/researchdesigns
14. Ho Yu, C. (2013, August 30). A Simple Guide to the Item
Response Theory (IRT) and Rasch Modeling.

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: A11500681S419/19©BEIESP 818 & Sciences Publication