Figure 1. System Architecture for Personalised Mobile Learning System
The increasing development in the world of Information Technology (IT) has provided a better way for teaching and learning. Learners have now preferred methods by which they learn and remember what they have learned with the use of mobile devices anywhere and anytime. A model widely used to determine learners' preference has been the Felder and Silverman learning style, but this learning style has been criticized due to its limitation of dichotomous responses in the Index Learning Style (ILS) questionnaire. Hence, this research paper developed a personalised mobile learning system (PMLS) that combines Felder and Silverman learning model with fuzzy logic to identify individual learner. In this study, the dichotomous ILS questionnaire is extended from the standard two (2) option ILS questionnaire to a five (5) option questionnaire in order to accommodate learners whose attributes fall in different dimensions. Fuzzy logic is applied to determine the degree of learner's preference or learning style. The system is implemented on Android based mobile devices. Experimental control group is used to ascertain the effect of learning preference in students' learning performance. The study used twenty-two participants of first year undergraduate students from Nigeria College of Education in 2018/2019 academic session. These set of students gainfully interacted with PMLS and also engaged in a learning system that is not PMLS (not focussed on personalisation). Results from the experimental control group showed significant improvement in students' learning performance from 50.95% (without PMLS) to 81.36% (with PMLS).
The increasing development in the world of information and technology (IT) has created new opportunities in the educational system. This has brought about modification in the traditional method of teaching and learning, causing a gradual shift of focus from the instructor who knows everything to students having some input and sharing their own knowledge in addition to the instructor's findings (Noor- Ul-Amin, 2013). This is not to downplay the role of the traditional mode of learning because it allows teachers to convey a lot of information to students and is a useful strategy for recall or rote learning (Agbonifo et al., 2013). However, it is an ineffective means to help students develop their mental capabilities as students with more cognitive capability tend to perform better than others. Modern technologies allow learners to learn and share information in an electronic format. E-learning has been described as a term generally used to refer to as computer enhanced learning (Adewale, 2007).
The quest to access information seamlessly from any point and in any place has made mobile devices an emerging technology with the potential to facilitate teaching and learning strategies that exploit individual learner's context (Sampson et al., 2012). This has brought about the concept of mobile learning popularly known as m-learning, which involves the process of learning and teaching with the use of mobile devices. It provides flexible on-demand access (without the time and device constraints) to learning resources, experts, peers, and learning services from any place (Kukulska-Hulme, 2009; Sampson et al., 2012; Traxler, 2009). In Adewale (2007), personalised learning is defined as the fitting of curriculum and learning support to meet the need of specific learners. This assists each student to learn conveniently in their own way by building a profile of each student's strengths and weaknesses and pace of learning. These differences in learning have been distinguished by psychologists as individual learning styles (Klašnja- Milićevićet al., 2011). The term learning styles refers to the concept that individuals differ regarding what mode of instruction or study is the most effective for them (Pashleret al., 2009). Common learning style models such as Felder and Silverman, Kolb, Mumford and Honey, Pask and more pointed that learners have different ways in which they prefer to learn (Klašnja-Milićevićet al., 2011). Although, some models are based on personality types while some are based on brain dominance, learners have shown to exhibit some characteristics that are common in some models which made it difficult to appropriately choose a model. However, researchers have further integrated multiple learning models while some combine learning style models with different techniques such as machine learning, fuzzy logic technique to achieve an optimal learning path for the individual learner.
Recently, there are many approaches to the development of learning tools for effective learning (Almirall et al., 2010; El-Bishouty et al., 2019; Guminska & Madejski, 2007; Huang et al., 2006; Martono & Nurhayati, 2014; Pajarito & Feria, 2015; Paulin et al., 2014; Saryar et al., 2019; Sprock et al., 2016). Most of these approaches use genetic algorithm, case-based reasoning, rule base or fuzzy techniques and more. However, learning in all these approaches does not optimally adapt to the learners' learning style or preference, nor implement on a mobile platform.
Hence, this research paper developed a personalised adaptive mobile learning system which delivers learning content to individual learner according to learner's requirements and capabilities as well as the degree of content visualization in learner's mobile device. Furthermore, Felder and Silverman Index Learning Style model and fuzzy logic technique are used to strengthen personalisation for a specific learner.
This section consists of the system architecture and system model.
The architecture of the personalised mobile learning system is a shown in Figure 1. The system is divided into two main parts namely, device part and server part. The details are explained in the following passages.
Figure 1. System Architecture for Personalised Mobile Learning System
The device part which is also known as the user interface layer enables learners to interact with all components of the system on their mobile devices. At the instance of network connections via wired or wireless technology, a new user registers by supplying all necessary information such as full name, username, email address, student identification number and password in order to uniquely indentify the users. As soon as this is completed, the profile user created by the system. User can subsequently log into the system through authentication process. It is through this interface that the user responds to questionnaires to determine his learning style.
All other operations that take place in the user interface include view and learning of recommended material, take post-test and rate the system. The goal of the user interface is to allow effective control and operation of the system.
The server part, also known as the back-end layer is made up of the application server and database server. The application server has direct interaction with the database server. The application server includes the fuzzy ILS model, fuzzy rule and device component content adaptation.
Fuzzy ILS model is used to generate the fuzzy ILS score from the ILS questionnaire responded by the user, It calculates the scoring based on Felder and Silverman method of generating ILS score. This score is used to identify learning style of learner out of eight learning style available which are active, reflective, intuitive, sensing, visual, verbal, sequential and global learning style and the degree of preference such as low, moderate and strong.
Fuzzy rules are used within fuzzy logic systems to infer an output based on input variables of learning style and its degree of preference identified by the Fuzzy ILS model. It makes use of IF THEN statement to match the learning style with appropriate learning contents from the database.
Device component content adaptation is used to format multimedia content available from the content database to adapt to the different screen sizes of mobile devices suitable for the user while also maximizing the user's quality of experience.
The database server is used to store all databases needed in the system. These include the Questionnaire database, profile database and learning content repository.
The personalised mobile learning system model is adapted from Felder and Silverman model approach combining with a fuzzy rule base technique. The system model is defined as:
where S represents the system, Cj=1,2,3,.., k denotes the set of course contents, Lp=1,2,3,..., w represents the set of learners and Di=1,2,3,.. z represents the mobile devices types such as Tablet, Personal Digital Assistant (PDA), smartphone, and more, and k, w, z are the total number of course contents, learners and mobile devices respectively.
Each device Di, of device component, contain features which are presented in equation (2) as:
where f1 represents bandwidth, f2 represents Processor speed, and f3 represents the screen size of device type. The course content format is in the form of text, video, audio, pdf and image is represented in equation (3) as:
where F is the course contents which in the form of outlines, animation, exercise, lecture delivery charts while text, video, audio, presentation are the format.
Felder and Silverman Learning model is used to classify learner into four Dimension as shown in equation (4):
where A represents Active, R represents reflective, Sn denotes sensing, I represents intuitive Vs represents visual Vb represent verbal Sq represent sequential, and G represents Global.
The computation of Felder-Silverman Fuzzy ILS scores is shown in equation (5) as:
where i(Dim, q) indicates the index i of a given pair of dimension, qend(A) represents the value produced from answering the questions in Dim +, and qend(B) represents the value produced from answering the questions in Dim -
The chosen values are further evaluated in their respective sets using the triangular membership function adapted from (Sproack et al., 2016) as shown in equation (6).
where x represents the total point of either end (A) or end (B) of the 4 dimensions. The result value from equation (5) is assigned to variable x in the membership function to determine the interval respectively. The interval indicates whether the learner's preference for an end in a dimension is strong, moderate, or low.
The formula in equation (6) indicates that when the values are within the range of 0 and 2 values, it is considered as Low. If the values are within the range of 2 and 9, it is considered Moderate, and finally, if the values fall in a range of 9 and 11, then Strong values within the end are evaluated. The preference low, moderate and strong of the function describes the linguistic terms as in equation (7).
The sample fuzzy rule is denoted as shown in equation (8) to determine the learning content F.
In designing the learning content, it is essential to take into consideration the characteristics of the learner in learning (Bajraktarevic, 2003). For efficient creation and order of learning content to the learner, Ahmad and Shamsuddin, (2008) method were adopted in this system due to the success of the research. Among the resource contents provided in the learning systems are as follows:
Animation: Successive drawings that create an illusion of movement when shown in sequence. The animations visually and dynamically present concepts, models, processes, and/or phenomena in space or time. Animations typically do not contain real people, places or things in movement. The activity is useful for Visual Learner.
Quiz/Test: Any assessment device intended to evaluate the knowledge and/or skills of learners. It is an online quiz that consists of multiple-choice questions and marks that can be displayed immediately after the student submits the quiz.
Hypertext: The learning content is composed of theory and concepts. Provide objectives, sub-modules, and navigation links. These learning resources are useful for Sensing/Intuitive students as well as Sequential/Global students.
PowerPoint Slide Show: Consist of example in the form of text, pictures and animations.
Exercises: Designed in multiple choice questions which students can answer and get hint and feedback regarding their performance.
Notes: Relevant text content related to the course.
Chart: Picture or Diagram or chart about a concept.
Lecture: An audio presentation of course or concept.
Open Textbook: An online textbook offered by its author(s) with its link.
Outline: Summary of course contents
Fuzzy ILS is computed from questionnaires responded to by the learner, each option has a score assigned to it for each pair of dimensions. Option “a” is 1 point for Dim + and 0 point for Dim , Option “b” is 0.75 point for Dim + and 0.25 point for Dim , Option “c” is 0.5 point for Dim + and 0.5 point for Dim , Option “d” is 0.75 point for Dim + and 0.25 point for Dim and Option “e” is 0 point for Dim + and 1 point for Dim . Table 1 shows how the system computes the ILS score for the entire Dimension.
Table 1. Adapted Learning Contents Developed Based on Felder Silverman Learning Dimension Characteristic (Ahmad & Shamsuddin, 2008)
The total point of the Dim + which is Active is shown as 7.25 while Dim which is reflective is 3.75. The points 7.25 and 3.75 are assigned to variable x in the membership function of equation (9) and (10) to determine the intervals respectively. However, since the value is with the range specified for Moderate, the values 7.25 and 3.75 are substitute for x into the equation (9) and (10).
For Active:
For Reflective:
The result of the calculation shows that learner has a moderate degree membership with a value of 0.78 to the whole Active membership, and a degree of membership of 0.34 Moderate in Reflective.
This same procedure is used to calculate for Sensing / Intuitive dimension, Visual/Verbal dimension and Sequential/Global dimension respectively in order to know the degree of membership preference for each learning style of each dimension. The result obtained shows that learner has a moderate degree membership with a value of 0.93 to the whole Sensing, and a degree of membership of 0.19 Moderate in Intuitive, 0.66-degree membership to visual and 0.47-degree membership to verbal and a strong degree membership with a value of 1 to the whole Sequential, and 0.13 value moderate degree of membership for global. These scores are used to decide which dimension that is preferable and which learning contents should be assigned based on the preference From these scores, the fuzzy rule is used to recommend learning content to the learner on a mobile device.
In this paper, fuzzy rules rule manually formulated based on selected input variables obtained from the ILS model and review of the literature. A rule is formed by extracting all the learning styles and degrees of preference. The learning style is Active, Reflective, Sensing, Intuitive, Visual, Verbal, Sequential and Global. The degree of preference are Low, Moderate and Strong. Some example of fuzzy rule use in developed PMLS include:
RULE 1: IF Active = Strong AND Sensing = Strong AND Sequential = Strong AND Visual = Strong, THEN Recommend Animation, Narrative text, Lecture Delivery, Examples, Diagram.
RULE 2: IF Active = Strong AND Sensing = Moderate AND Sequential = Strong AND Visual = Moderate THEN Recommend Exercise, experiment/practical, PowerPoint slide.
RULE 3: IF Active = Strong AND Sensing = Strong AND Sequential = Moderate AND Visual = Moderate THEN Recommend Animation, Exercise, concrete content (hypertext).
RULE 4: IF Active = Strong AND Sensing = Moderate AND Sequential = Moderate AND Visual = Strong THEN Animation, Narrative text, Diagram.
RULE 5: IF Active = Moderate AND Sensing = Strong AND Sequential = Strong AND Visual = Moderate THEN Recommend Exercise, concrete content (hypertext), lecture delivery.
The system flowchart is depicted in Figure 2 shows how the learner is allowed to create a profile if it is not created. In creating a profile, the learner supplies his identities such as student ID, name, level, email address, username and password. As soon as the profile is created, a learner can log in and authenticated and have access to answer to questionnaires. The ILS score is generated and display, the highest score among the ILS score displayed is used to identify the learning style of the learner. The fuzzy rule is applied to match learning content in database to learning style, learning contentment adapt to learner's mobile device for the learning process. If the content is not matched with the appropriate learning style, the learner is redirected. The learner profile, learning style is updated in the database. After completion of learning, the learner takes survey questions and post-test to evaluate system performance and learner performance respectively.
Figure 2. System Flowchart of PMLS
Table 2. Fuzzy ILS Score used in Determining the Learning Style of a Learner
Twenty two (22) students studying Information Technology in Umar Suleiman College of Education, Gashua, Yobe State, Nigeria are used for test cases in this research paper. They are first-year undergraduate students, male and female between the ages of seventeen (17) years and twenty-four (24) years old. The students actively interacted with PMLS and the assessment component (named Post-Test) in PMLS; furthermore, they also engaged with another learning system that is not PMLS and its assessment component (named Pre-Test) for the purpose of affirming the significance of learning preference in students' learning performance. Tables 3, 4 and 5 show the Pre-test score, Post-Test score and percentages of students performance in both Pre-Test and Post-Test scores respectively. The outcome from these results shows that only 2 students score 70 % and above in the Pre-Test which is 9% while 17 students score 70% and above in the Post-test.
Table 3. Students' Performance in Pre-Test
Table 4. Students' Performance in Post-Test
Table 5. Analysis of Students' Scores in Pre-Test and Post Test
In the Pre-Test score, 9% of students passed with A, 27% of students passed with B, 9% of students passed with C, 37% passed with D and 18% failed as depicted in Figure 3. Figure 4 shows Post-Test score with 77.3% students passed with A, 18.2% passed with B and 4.5% passed with C. The poor performance in Pre-Test is due to not incorporating the learning strategy and approach that address the distinct learning needs, interests, aspirations, or cultural backgrounds of individual students. On the other hand, the accelerated performance in Post-Test is due to the use of personalised learning system and mobility of the system that allow students to learn anywhere and anytime. Figure 5 shows the comparison between Pre-Test and Post-Test results.
Figure 3. Percentage of Students' Performance by Grade in Pre-Test
Figure 4. Percentage of Students' Performance by Grade in post-Test
Figure 5. Comparison of Students' Scores in the Pre-Test and Post-Test
The performance evaluation of the system is carried out using a survey (questionnaire) which focuses on the following metrics such as user interface design, navigation, ease of learning, technical accuracy, ease of operation, efficiency and learning experience. The survey uses five point-likert scale with the following linguistics terms and values: very poor (1), poor (2), average (3), good (4) and excellent (5). There are twenty-two (22) participants (students) that responded to the questionnaire and their responses are analysed using weighted mean average as depicted in Table 6. The weighted mean is calculated using the formula in equation (11).
Table 6. Analysis of Survey's Respondents by Weight Mean
where Σ denotes the sum, w is the weights and x is the value. The graphical representation of the weighted mean analysis is shown in Figure 6.
Figure 6. Graphical Representation of Weighted Mean Analysis
Comparing PMLS with existing systems developed by Martono and Nurhayati (2014); Pajarito and Feria (2015) and El-Bishouty et al. (2019) showed the following findings from the review of literature: Martono and Nurhayati (2014) developed an android based mobile learning application as flexible learning media but the learning contents are not personalised. Pajarito and Feria (2015) developed a platform that implements the standards by creating micro-learning contents for mobile devices. The creations of microlearning reduces a learning content to its smallest form for easier transmission and lower storage requirements. However, the system did not take into account the learners profiles and preferences on a mobile platform. El-Bishouty et al. (2019) focused on the use of Felder and Silverman learning style model to design online course to identify the learning style of the students. The dichotomy in options of Felder and Silverman ILS questionnaire did not adequately capture learner's characteristics for effective learning content recommendations and the system is not evaluated to ascertain adequate recommendation of learning contents and the usability of the model. Hence, PMLS addressed the limitation of dichotomous responses of Fielder and Silverman ILS questionnaire by extending the options to five likert scale which provide adequate capturing of learning style and preference of students for a suitable learning contents on mobile platform.
The use of mobile devices to enable new ways of learning has been an area of interest in research for quite some time. Mobile devices have the chance to foster personalised learning activities in spite of time and location. The learning processes are more effective and produce the desired result if learning is tailored to address the distinct learning needs, interests and cultural backgrounds of individual students. In this research paper, personalised mobile learning system using the fuzzy logic technique with the emphasis on the learning style is developed. The Index Learning Style of Felder and Silverman style model questionnaires with five (5) likert scale options instead of two (2) options that suffer dichotomous is used to determine individual learning style. The learner is categorised into (8) learning styles such as active learner, reflective learner, sensing learner, intuitive learner, sequential learner and global learner. Each learning style has a degree of its preference such as strong preference, moderate preference, and low preference respectively. Learning contents are recommended to the learners based on their learning style preference using the fuzzy logic rule. The study used twenty two participants of first year students 2018/2019 from Nigeria college of Education to attest the significance of learning style and preference on students' learning performance. The students participated both in PMLS with its assessment component (Post-Test) and another learning system that is not PMLS with its assessment acomponent (Pre-Test) for a certain period of time. The results from both Pre-Test and Post-Test showed an increase in students' learning performance from 50.95% (not PMLSl) to 81.36% (with PMLS). Future work could be done by integrating context-aware component into the system in order to deliver appropriate learning contents to students based on various contextual information. Furthermore, Ant Colony Optimisation (ACO) approach could be incorporated to provide selection of an optimal learning path suitable for students.