Dynamic Knowledge Tracing


Knowledge tracing refers to the problem of modelling knowledge of students over time as they interact with a learning system, which have been studied extensively in computer supported education. However, the knowledge tracing problem is inherently difficult as human learning is grounded in the complexity of both the human brain and human knowledge. Recently, deep neural networks have been successfully used to build several knowledge tracing models, including the Deep Knowledge Tracing (DKT) model developed by the researchers from Stanford University (Piech et al., Deep knowledge tracing, NIPS 2015) and Dynamic Key-Value Memory Network (DKVMN) model ( Zhang et al., Dynamic key-value memory networks for knowledge tracing, WWW 2017),  which have been shown to be able to predict student performance better than the previous knowledge tracing models.



The objective of this research project is to investigate the capabilities of various deep learning
models for knowledge tracing to predict students learning behaviour, aiming to extend these models
to provide real-time personalized learning recommendations for students.


Strong skills in programming (C++, Java or Python) are required. Solid knowledge background in data mining and machine learning are desired.


Background Literature

  • Piech et al., Deep knowledge tracing. In Advances in Neural Information Processing Systems, 2015
  • Xiong et al., Going deeper with deep knowledge tracing. In Educational Data Mining, 2016
  • Zhang et al., Dynamic key-value memory networks for knowledge tracing. In Proceedings of the 26th International Conference on World Wide Web, 2017


Gain a solid understanding of deep learning models, and learn how to implement and apply these techniques into a real-world application setting.



Machine learning, knowledge graphs, artificial intelligence, data mining


Updated:  1 August 2018/Responsible Officer:  Dean, CECS/Page Contact:  CECS Marketing