Data-driven modelling of learner's cognitive style in educational hypermedia

摘要

In this paper we propose a framework for modelling the user behaviour in hypermedia systems. This involves the design of time-based features and the selection of the most useful ones that can give the best classification and prediction. The process of variable selection involves a sensitivity analysis via neural network bootstrapping, which aims at maximising the model's classification performance and generalisation ability. The goal of this study is to model and assess the learners' holist/analytic cognitive styles based on the navigational trail recorded while they navigate through learning hypermedia content. The method is generic in nature and therefore applicable to a wide range of behaviour-analysis applications.