Is Deep Knowledge Tracing Using Recurrent Neural Networks Effective at Improving Online Educational Platforms?
The extensive usage of online learning platforms like Moodle among university students creates a vast source of users' behaviour data. Analysing this data with a machine learning model called Deep Knowledge Tracing (DKT) and accordingly adjusting the content and style of online materials can be beneficial for both students and educators.
DKT can detect the pattern of learning actions by analysing data from past student actions. The model can find the most effective way a student learns- for example by practising more exercises or reading additional literature. Since DKT is closely tailored to predict studying behaviour, the model can be incorporated into online educational platforms that automatically adjust the content and materials depending on the needs of the user. The benefit is that a large group of students with different learning methods can make the most out of an online course due to the flexible content generated.
From a lecturer’s viewpoint, DKT using Recurrent Neural Networks (an algorithm for learning contextualised data by repeatedly extracting hidden information) can produce data on where students have certain difficulties with a course. By exploring the platform usage, data models can discover the level of user's confusion. The analysed data can be used to inform educators on students' difficulties and allow them to provide just-in-time support. Having up-to-date information on students' struggles allows lecturers to adjust the content of a course, decreasing dropout rates.
This paper explores the effectiveness of DKT for educational platforms. Consequently, the benefits, drawbacks, and the practical applications for all participants in the learning process are presented in the form of research evidence.