Decision-Making Model in E-Learning - Coursework Example

Proposal for a decision-making model in E-learning by A proposal for a decision-making model in e-learning using an original algorithm which academic institutions can use to select the best LMS-Learning Management system that will enable it achieve its goals. The decision made should be on the basis of a certain criteria weighted using a mathematical model. The performance of the novel algorithm proposed is also performed. Abstract More and more digital learning resources are available online through various learning management systems; this has made it a challenging task for learning institutions to choose the appropriate learning management system that will enable its achieve its goals as well the learning goals of its learner. Therefore, there is a need for a new tool which would be able to recommend a more appropriate learning management system for an academic institution based on its needs and goals. This project therefore, proposes to implement an intelligent decision support system for recommending learning management systems or e-learning platforms based on a novel algorithm. The recommendations are on the basis of weighting criteria based on a mathematical model that analyses a user’s performance during the training process on one hand and the search queries made by the user on the other hand. This tool uses and applies k-means clustering algorithm and fuzzy logic to determine document arrangement based on their performance level. Further, the project proposes a new ontology domain alignment technique that uses contextual data of the knowledge sources for decision making from e-learning domain. This domain has been tested empirically, and the results show that it performs better than other existing methods. This paper makes salient contributions such as the use of k-means clustering algorithm for decision-making and fuzzy approach for ontology alignment. Introduction The internet has had significant impact on the establishment and development of e-learning education. The evolution of internet technology as well as e-commerce has also affected all commercial and industrial activities which have in turn resulted in accelerated growth in the e-learning industry (Tzeng et al. 2007). The advancements in internet technology have also supported the development of e-learning by enhancing the speed of transfer and volume of information, as well as simplified exchange of tasks and knowledge management. An e-learning platform is an emerging academic delivery tool for most academic institution, which have today developed their own e-learning course especially targeting on-job training person interested in advancing their education without physically attending college, via the internet It is important to evaluate the effectiveness and appropriateness of e-learning systems despite the fact that they have developed over time. There are a myriad of studies which propose several factors that need to be considered in evaluating and determining the appropriateness of an e-learning or learning management system for an academic institution (A. J. Kay & Holden 2002). In fact, there are several decision-making or evaluation models out there, which are based on specific factors. These numerous evaluation model, apparently influence one another in one way or the other. These decision-making models, however, are in some way deficient since they do not have guidelines for appropriateness and effectiveness evaluation. It is of great importance for any proposed appropriateness and effectiveness evaluation to incorporate and consider the goals of an academic institution on the basis of learning theories, relative user-interface design, as well as users learning satisfaction (Tzeng et al. 2007). Since different criteria and aspects can be used for the evaluation of e-learning platforms, a suitable approach would be the use of MCDM-multi-criteria decision making model (Wang 2003). E-learning, according to Azzeh et al. (2008)has become a non-linear, easily accessible learning process widely applied in dynamic and distributed environments such as on cloud technologies and on the internet. The main advantage of e-learning technology today is the fact that it facilitates adaptive learning. This enables tutors, lecturers, and teachers to dynamically deliver and revise instructional materials on the basis of the progress of learners. Otherwise, tutors, lecturers, and teachers are forced to peruse and revise and deliver the documents manually, a process that is cumbersome and time consuming. This challenge can only be overcome by using an automatically built ontology. Ontology plays a crucial role in seizing and distributing the idea for effective and efficient HCI-human computer interaction. It should however, be noted that domain ontology engineering is just as cumbersome and time consuming as the manual process. Various researchers have in past examined various machine learning methods for semi and automatic domain ontology discoveries. These methods, however, were found to be insufficient in their computational accuracy and efficiency for effective decision-making with regards to e-learning platforms (Tzeng et al. 2007; Wang 2003; Perera et al. 2009). A critical issue for e-learning platforms is their ability to monitor the progress of students’ learning performances. This paper, therefore, aims at assisting institutions, specifically academic institutions to determine automatically a fast way for the construction of ontology for effective decision-making, specifically from large collections of document sources. E-learning technologies or learning management systems technologies can automatically support an automatic method for constructing concept maps and for the analysis of learners in order to determine and characterize learners understanding of the topics that they learn via an e-learning platform. This work implements a concept map generation mechanism which is supported by a fuzzy domain ontology mining and a situation-sensitive text extraction (R. Lau et al. 2009; Tho et al. 2006; Pinto & Martins 2004; A. J. Kay & Holden 2002). Concept maps, according to are important in the generation of ontology; in fact, ontology offers and efficient concept representation and the representation of semantic associations among concepts. This paper has two main goals; one is to develop original indistinct domain ontology mining method. It is worth noting that, this is despite the fact that there are a myriad of many other learning techniques out there that have been previously proposed for semi-automatic and automatic mining of domain ontology. However, the fuzzy domain ontology being proposed in this paper utilizes intelligent techniques in order to improve the efficiency, reliability, and accuracy of decision making. Apparently, all the methods proposed and researched in the past are still lacking in learning accuracy and computational efficiency and reliability. Ideas have been recognized by researchers such as the most significant asset in the e-learning scenario; it is in fact, considered as a major component in e-learning platforms for understanding a given subject designed and implemented. A tutor, lecturer, or a teacher can perform adaptive learning and teaching on the basis of the knowledge structures of his or her learners, in order to improve the achievement of e-learning goal. One of those features used to represent the knowledge structures of a particular application domain is domain ontology. A good example is the formal and explicit capturing of knowledge regarding markets, people, environment or businesses in such a way that it can be shared among computer systems and humans. Many learning management systems or e-learning platforms are being developed with suitable techniques that are able to capture and epitomize appropriate and suitable knowledge. In this regard, researchers such as Miller et al. are also working on developing suitable techniques that are able to capture and epitomize appropriate and suitable knowledge. Miller et al. proposed a data extraction approach that would help discover Meta data relations that define different learning resources. Terms from meta-data files were, in this study, parsed, while stop words were removed. The author applied tools such as Word Net to extract word roots; on the part of tagging speech, an algorithm, the Brill tagger, was used. In other studies, (Perera et al. 2009; R. Lau et al. 2009), IDF-Inverse Document Frequency and TF-Term Frequency heuristic for Information Retrieval are used to mine e-learning electronic generated messages. The knowledge density score in such systems were determined by the IDF/TF term weighting formula for assessing the extent to which the system contributes to the sharing of information online by individuals. Clustering was applied in A. J. Kay & Holden (2002)to determine and find out groups of similar members and teams. Data was mined by collecting students’ information from student blogs and online discussion boards. In such a scenario, the k-means clustering technique can be utilized for documentations that may require very few clusters; this would help the set of words that are pre-selected. Therefore, this paper employs speech tagging based on WordNet in addition to utilizing the document parsing technique that uses IDF/TF, among other language pattern recognition techniques of mining concepts from texts. Additionally, this paper also addresses the concept of automatic development of classification concepts. Thresholds capable of measuring students understanding levels are used. Concept clustering and similarity measuring technique, particularly the Jaccard Similarity technique is employed to put together, into set up groups similar documents using the k-mean clustering. Rules and agents are also used in this paper to provide suitable in put on the basis of the built ontology. While building ontology, fuzzy logic is used for classification. Intelligent decision making technique algorithm software is proposed for determining the effectiveness and appropriateness of an e-learning platform or technology, as well as for enhancing and showing how fuzzy domain ontology mining algorithm and the context-sensitive text extraction technique can be applied to generate concept maps automatically in order to e-learning student’s knowledge structures (Perera et al. 2009; Azzeh et al. 2008). Therefore, academic institutions can be able to choose appropriate learning management tools or e-learning platform that would meet their goals based on the information that the ontology maps will disclose. This proposed decision-making model in e-learning makes major contributions in terms of its design of a new architecture for concept generation, similarity grouping and computation for identifying the right and relevant course content, and for pre-processing indexing. Proposed decision-making model The proposed decision-making model proposes a system architecture which is interactive, collaborative, intelligent, distributed, and adaptable E-learning platform. This is a system used to retrieve relevant information with regards to the learners as well as support instructional design. It analyses and processes the data results which thus enables significant recommendations for e-learning. Today, most e-learning platforms or learning management systems only put into consideration one feature. Thus, it is important for academic institutions to select e-learning systems that support all the features that would otherwise be considered appropriate and efficient. Learners Knowledge Pre-processing Basically, this is made up of a process of optimizing the list of terms which identify a taxonomy or collection. This module is used receive text inputs from the text corpora. The text file is then, using a tokeniser, converted into tokens, each of which is later passed on to the stop word remover where words such as prepositions and their determiners are removed. This is because stop words appear in any given context and do not offer any significant information to describe a domain idea (R. Lau et al. 2009). The implementation of the proposed decision-making model, the construction of stop word file is based on a standard word file. In fact, after the stop word removal, the words obtained are stored in different text file, which is then read and passed on to the process of POS tagging. Learners Knowledge Raw Term Generation and Indexing Raw term and indexing generation receives its input as tokens, which are divided into various documents from the pre-processing modules. Each of the documents is made up of about a set of ten or five terms; the occurrences of each term in the documents are computed. In order to show the terms in columns and rows for the document, a matric is generated; an index is also used to arrange the terms in descending and ascending order. Concept Filtering of the Knowledge of Learners IDF and TF concept filtering is performed. It is very possible to use TF indexing to normalize the raw frequencies in any given document. For instance, a document with two words, one of which appear in the document thrice while the other appearing in the document four times, the first word will be normalized to 0.4 or 3/7, while the other will be normalized to 0.6 or 4/7. Thus, the number of times a term appear in a document is referred to as the term count, which is usually normalized in order to prevent bias and to give a term t some measure of importance with a given document d. the term frequency TF used is therefore TF(t, d). IDF-Inverse Document Frequency according to [6] tries to smoothen the frequency of a given term within documents. It considers a word or term appearing in more than one document as one with less precision and whose value should go down. Therefore, the weight of concept is calculated by combining IDF and TF. Consider: In this equation, D refers to the total number of documents in corpus, refers to the number of documents in which the term t appears. In circumstances where the term is not within the corpus, there will be a division by zero, as such, the formula can be adjusted to. is then calculated using the formula Apparently, a high weight is reached when there is a low document frequency and high term frequency in a whole document collection; such weights therefore attempt to filter out some common words. Indexing Latent Semantics According to R. Y. K. Lau (2003), indexing latent semantics tries to determine the latent relationships that exist among documents on the basis of the occurrence of terms. Thus, if document X contains, and document Y contains, it can be concluded that a common latent relationship exists between documents X and Y; for instance, is common among these documents. In this case, X represents the novel matrix, B is the sigma vector, A the word vector, and C represents the document vector. By using SVD-Singular Vector decomposition, LSI-Latent Semantic Indexing gets the different matrices, A, B, and C.; these are then reduced and rebuilt to form the original vector. In this regard, noisy relations are suppressed resulting in clearly visible relations. The Computation of Similarity Based on mutual data, similarity computation is performed. Mutual data refers to a theoretic information technique used to compute and determine the existence of dependency two entities (Pinto & Martins 2004). Similarity computation can also be carried out using a technique referred to as BMI-Balanced Mutual Information. This computes and determines the weight of association among tokens (R. Y. K. Lau 2003). As had earlier been mention, this proposal utilizes Jaccard computational method. This is useful in computing the weight of association among tokens, and it puts into consideration both presence and absence among documents as the associations. This is important in estimating membership values with regards to a concept. They are computed and determined for attribute, relation, and the definition of concept in fuzzy domain ontology. Using Jaccard’s coefficient, the similarity measure is given by: In this similarity measure formula, the Jaccard’s coefficient is used to determine the measure of similarity just as it is used in many other systems which utilize this formula to determine representation in ontology model for machine learning methods. The proposed model in this paper maps the similarity between two ontologies onto one new ontology [5]. Using this coefficient, similarity probabilities measure of the structure of ontology of each document can be computed. Similarity Computation using Fuzzy Logic [4] proposes the computation of similarity measure on the basis of fuzzy logic; specifically for measuring the similarity of software project. This will be ideal, especially for decision making by academic institution in determining the effectiveness and appropriateness of e-learning platforms. This paper uses fuzzy rules to handle the uncertainties regarding the prediction of learners learning styles. These are styles that are determined based on the identified learners’ membership values based on categories including low, high and medium, which are utilized in classifying fuzzy ontology. Consider Figure 1; a similarity fuzzy relation model, which accepts 0.5 as a threshold for both high and medium similarity levels. It considers a student to be of a low similarity level if he is taught for a myriad of times with better learning methods and materials; medium similarity is when the learning materials used contains more explanations, while high similarity level is where learning materials are sent to a knowledge base where a student can access and use it in future. Figure 0‑1: Similarity fuzzy relation model The symmetric fuzzy logic model described in this proposed model identifies three different categories of learners on the basis of their learning styles. Accuracy percentage is proposed to be used in evaluating this model so long as the total number of documents sampled is considered. It is expected that in comparison to existing algorithms such as Bayesian algorithm, the proposed model will have increased percentage accuracy. This is as a result of the fuzzy inference engine being used being effective and sufficient in handling information somewhat considered uncertain. Clustering is also used to convert large amounts of documents into similar learner behaviour groups. This is done through the use of the k-means clustering algorithm, which form efficient groups of similar documents. A new procedure is proposed which efficiently operates to producing similar documents classification in high dimensions. On the basis of the clustering results, learning materials on similar subjects are either grouped as medium, simple, or hard, based on understanding. This grouping is utilized by the decision model to provide learner with appropriate and suitable materials. Conclusion and Recommendations Evaluation of e-learning is still not sufficient since no evaluation guidelines exist. There are a myriad of intertwined criteria and facets that could be used to evaluate e-learning platforms; most of them, however, are not efficient or sufficient. Therefore, a multi-faceted decision making model would appropriate, suitable, and sufficient for evaluating e-learning platforms. This paper proposes an intelligent decision making MCDM model which generates concept map ontology construction and decision-making in e-learning. The design and implementation utilizes intelligent rules and integrates a myriad of techniques. Learning materials are classified and present to the users depending on the constructed fuzzy ontologies. Further, documents from learners are classified on the basis of fuzzy logic ontology and similarity is computed using Jaccard’s similarity, fuzzy set, as well as clustering methods and techniques. Apparently, this yields more efficient results to examine the quality, relevance, and appropriateness of the extracted document. Based on this model, lecturers, instructors, and teachers can be able to effectively and quickly determine their students’ learning status and progress. Fuzzy ontology is considered as an appropriate and effective way for identifying and determining documents ontology similarities, an important feature in e-learning decision making. The model proposed in this study would be able to evaluate the appropriateness and suitability of an e-learning platform, in addition to displaying the association of the intertwined facets. Therefore, if an e-learning platform is deemed deficient, ineffective or inappropriate, the model enables the determination, based on AHP weights and on impact-digraph-map, of the problem and deficiencies of the e-learning platform. Bibliography Azzeh, M., Neagu, D. & Cowling, P., 2008. Software Project Similarity Measurement Based on Fuzzy C-Means, Berlin: Springer. Kay, A.J. & Holden, S., 2002. Automatic Extraction of Ontologies from Teaching Document Metadata. In Proceedings of the 2002 International Conference on Computers in Education. pp. 1555–1556. Lau, R. et al., 2009. Towards a Fuzzy Domain Ontology Extraction Method for Adaptive e-Learning. IEEE Transactions on Knowledge and Data, 21(6), pp.800–813. Lau, R.Y.K., 2003. ContextSensitive Text Mining and Belief Revision for Intelligent Information. Web Intelligence and Agent Systems An International Journal, 1(34), pp.1–22. Perera, D. et al., 2009. Clustering and Sequential Pattern Mining of Online Collaborative Learning Data. IEEE Transactions on Knowledge and Data, 21(6), pp.1–14. Pinto, H.S. & Martins, J.P., 2004. Ontologies: How can they be built? Knowledge and Information Systems, 6(4), pp.441–446. Tho, Q.T. et al., 2006. Automatic Fuzzy Ontology Generation for Semantic Web. IEEE Transactions on Knowledge and Data Engineering, 18(6), pp.842–856. Tzeng, G.-H., Chiang, C.-H. & Li, C.-W., 2007. Evaluating intertwined effects in e-learning programs: A novel hybrid MCDM model based on factor analysis and DEMATEL. Expert Systems with Applications, 32, pp.1028–1044. Wang, Y.C., 2003. Assessment of learner satisfaction with asynchronous electronic learning systems. Information & Management, 41(2), pp.75–86. Read More