Development of Intelligent Tutoring Systems Using Hybrid Machine Learning Approach

Simple item page
AuthorAlshaikh, Fatema
metadata.dc.contributor.departmentUniversity of Bahrain, College of Information Technology - COMPUTER ENGINEERING -Theses and Dissertations
metadata.dc.date.accessioned2026-09-20T15:48:37Z
Publication date2024
Degree levelMaster
AbstractAbstract : The primary goals of research in the fields of education and Artificial Intelligence (AI) are to develop an autonomous agent that can perform tasks or make decisions without hu- man intervention. An Intelligent Tutoring System (ITS) is essential to education because it gives students individualized, personal assistance as they learn problem-solving techniques through instruction and immediate feedback. This thesis presents three experiments with different architectures to investigate the effectiveness of AI and machine learning algo- rithms in the domain of intelligent tutoring system. The first experiment builds three models based on the structure of Convolutional Neural Networks (CNN) to investigate the efficacy of Deep Learning (DL) algorithms in predict- ing student academic achievement. Two public datasets are used (PTdataset) and (xAPI- dataset). According to the experimental findings, the proposed CNN model's success in predicting student performance at early stages reached an accuracy of 94.36% using the PTdataset and 84.83% using the xAPIdataset. In the second experiment, we used the Learning Classifier System (LCS) of Genetic Algorithms (GAs) and Reinforcement Learning (RL) to develop an adaptive educational model and construct a learning environment that is personalized for learners. We developed three models. The first model utilized LCS, and the second model utilized RL. The third model integrated a learning classifier system using genetic algorithms with reinforcement learning to create a hybrid approach, which we named it Adaptive Pedagogical Model Based on Learning Classifier System and Reinforcement Learning (PMCR). The study's findings indicate that PMCR more ideal than the other proposed models. It achieved an accuracy of 82% and an F1 score of 0.9 when tested across various problem cases. The third experiment is based on Deep Reinforcement Learning (DRL). In ITS, DRL provides students with individualized learning experiences based on their own learning styles and preferences. We presented four different models. The first model is based on the structure of the Deep Q-Network (DQN). The second and third models (Double DQN and Dueling DQN) are extensions of DQN algorithm to address some limitations of the DQN approach. The last proposed model is a hybrid model that combines the standard DQN with a learning classifier system. The findings of the study indicate that both Double DQN and Dueling DQN showed complete data efficiency with 100% accuracy, 1.0 F1 score and 1.0 precision at highest training level of 1000 episodes and 10 steps whereas DQN performed fairly but was outperformed by advanced techniques.
metadata.dc.identifier.otherDrupal node 3751
metadata.dc.identifier.otherhttps://digitalrepository.uob.edu.bh/en/dar/development-intelligent-tutoring-systems-using-hybrid-machine-learning-approach
metadata.dc.identifier.urihttp://165.227.56.122/handle/123456789/6778
Languageen
Linkhttps://digitalrepository.uob.edu.bh/en/_flysystem/fedora/2025-07/00876_development_of_intelligent_tutoring_systems_using_hybrid_machine_learning_approa.pdf
TitleDevelopment of Intelligent Tutoring Systems Using Hybrid Machine Learning Approach
TypeMaster's Thesis

Files

Original bundle

Now showing 1 - 1 of 1
Thumbnail Image
Name:
node-3751-1-00876_development_of_intelligent_tutoring_systems_using_hybrid_machine_learning_approa.pdf
Size:
6.24 MB
Format:
Adobe Portable Document Format
Download

Collections

Master's Theses