An intelligent problem selection agent for SQL-Tutor using artificial neural networks
Timothy Wang
Department of Computer Science
University of Canterbury
Abstract
The use of Intelligent Tutoring Systems (ITS) is necessary to alleviate the teaching shortage that has effected educators in recent years. Structured Query Language Tutor (SQL-Tutor) is an ITS developed at the University of Canterbury and is used for teaching undergraduate database courses.Artificial Neural Networks (ANNs) is a Machine Learning approach that is a simple approximation of the human brain. ANNs are very good at learning in domains where there are no well defined algorithms or the domain is not well understood. The research presented investigates the possibilities of using ANNs for making pedagogical decisions.
The use of ANNs for this project focused on the selection of appropriate problems for students to work with. Two ANNs were used to select a suitable problem. The first network was designed to assess whether a student is struggling with a problem. If it has been determined that the student will have difficulty, the problem selector finds an appropriate problem. Prediction of whether the student will have difficulties with a problem achieved 93% accuracy. The second network selects the problem that is best suited to the student's level of ability. Prediction accuracy achieved with this network is on average 79%.
The first network performed well in assessing whether a student will have difficulty with a problem. The second network was less successful when finding an appropriate problem for a student to attempt. The results do suggest that there is a good basis to use ANNs with SQL-Tutor and other ITSs.
