A dynamic eLearning prediction modelbased on incomplete activities of eLearning system

• Main Author: Chayanukro, Songsakda
• Format: Thesis
• Language: English; English; English
• Published: 2020
• Subjects: T58.5-58.64 Information technology; L Education (General)
• Online Access: https://etd.uum.edu.my/9162/1/Deposit%20Permission_s93189.pdf; https://etd.uum.edu.my/9162/2/s93189_01.pdf; https://etd.uum.edu.my/9162/3/s93189_references.docx; https://etd.uum.edu.my/9162/
• Institution: Universiti Utara Malaysia

At present, eLearning usage is diverse because the eLearning activities used in teaching and learning differ depending on educators. The selection of activities in different eLearning usage affect the prediction of learning outcomes. However, most eLearning outcome prediction models are still unstable and inapplicable in many situations as the eLearning usage is considered to be highly dynamic. Therefore, the objectives of this study are: a) to analyze the eLearning activities that affect learning outcome; b) to construct a learning outcome prediction model for eLearning usage; c) to synthesize a dynamic eLearning prediction model based on incomplete activities of eLearning systems; and d) to evaluate the dynamic eLearning prediction model based on advantage, accuracy, and effectiveness. This study was conducted through seven steps: initial study; data collection; data pre-processing; eLearning activity analysis; learning outcome prediction model construction; eLearning prediction model synthesizing; and model evaluation. Six data mining algorithms were used in evaluating the model. The results found seven significant groups of eLearning activities that could predict the learning outcome with more than 75% accuracy. Of the seven significant groups, two groups of activities have Receiver Operating Characteristic values greater than 0.5. Hence, this study demonstrates that using data from incomplete activities of eLearning systems provides an appropriate means for predictable learning outcomes. The prediction model contributes to an optimal number of classes and data set where two classes received the highest accuracy ratio. Practically, the results of this study may assist towards improving management and reducing educational costs.