Adaptive e-learning using ECpAA rules, Bayesian networks and group profile and performance data

In order to deliver individualized instruction to learners, an adaptive e-Learning system must be able to perform content selection, sequencing and presentation, and to control learners' navigation through content based on their different profiles and performances. However, the profile data provided by learners and the performance data gathered by a system may be incomplete, inaccurate, or contradictory.;This dissertation presents techniques and features, which alleviate the above data problems by evaluating the profile and performance data of each new learner probabilistically based on the profile and performance data of earlier learners. We present the methodology for the construction, search, and selection of learning objects. Our probabilistic rule model allows our system to apply adaptation rules to examine learners' data at various stages of processing a learning activity, and make proper adaptation decisions even though the learners' data may have anomalies. Adaptation rules are processed by a rule engine and a Bayesian Model Processor. Content authors are provided with system facilities to define adaptation rules and Bayesian Models. The prior distribution of a Bayesian model is automatically derived by using the formulas presented in this work together with prior probabilities and weights assigned by the content author or the instructor. Each new learner's profile and performance data are used to update the prior distribution, which is then used to evaluate the next new learner. The system thus continues to improve the accuracy of learner evaluation as well as its adaptive capability.;Two applications have been developed to demonstrate several features of adaptation: namely, construction of self-contained and reusable learning objects, dynamic content search and selection, event and rule-based processing of learning objects at six adaptation points, probabilistic specification and evaluation of rule conditions using Bayesian Models, and the support of personalized learning depending on a learner's profile and performance as well as the characteristics of content. Several simulations have been conducted to handle data anomalies, and verify our proposed formulas for deriving the conditional probabilities needed for Bayesian inference. The system architecture and implementation of an adaptive e-learning system are also presented.