Student modelling and classification rules learning for educational resource prediction in a multiagent system

To model support for human learning, rules (i.e. triggering event-conditions-actions) can be classified to encompass any state of student learning activity enroute to appropriate learning material prediction. In an agent based system, each component of an adaptive multiagent system can be represented as agents having individual autonomy and responsibility to realise the overall goal of the system. In this paper, we present an extended work on a multiagent based Pre-assessment System in which a modelling agent employs the technique of One v All Multiple Classification rules to make predictions for learning materials after some prerequisite assessment facts to a desired concept or topic are communicated by the pre-assessment agent. Using SQL ontology tree structure as the domain of learning content, a learning algorithm is described as a process for estimating the total number of classified rules required for the pre-assessment system. This estimate is proven to be dependent on: 1) two binary state values, 2) the number of leaf-nodes in the ontology tree, and 3) the number of prerequisite concept(s) to a desired concept. In addition, is the learning algorithm with which a modelling agent can increment or decrement its classified number of rules.