Student Action Prediction for Automatic Tutoring for Procedural Training in 3D Virtual Environments

Abstract

This paper presents a way to predict student actions, by using student logs generated by a 3D virtual environment for procedural training. Each student log is categorized in a cluster based on the number of errors and the total time spent to complete the entire practice. For each cluster an extended automata is created, which allows us to generate more reliable predictions according to each student type. States of this extended automata represent the effect of a student correct or failed action. The most common behaviors can be predicted considering the sequences of more frequent actions. This is useful to anticipate common student errors, and this can help an Intelligent Tutoring System to generate feedback proactively.