Ph.D. in Computer Science, my advisors are Prof Neil Heffernan and Prof Joseph Beck. My research topic is educational data mining, and my work focuses on building an adaptive learning system called the Automatic Reassessment and Relearning System. The system is backed by well-studied and improved spacing effect theory and "actionable" machine learning models to schedule spaced-out learning sessions for individual students. Since 2012, over 35,000 students used my system and it yields reliable improvement on students' long-term performance. On the other hand, I have also built good publication records (14 papers, 2 submitted) in the areas of EDM and ITS, including topics on predictive models, randomized control trials, adaptive learning platform, and deep learning. Here are some links to showcase my work:
1) I built an adaptive learning system, called Automatic Reassessment and Relearning System (ARRS), for the ASSISTments platform to help improving students' retention performance. It has been featured in this paper and in my thesis.
2) I discovered a feature, called Mastery Speed, which has great predictive power for modeling retention performance. See this paper.
3) I built an ETL pipeline to support the research of advanced data mining and machine learning at ASSISTments lab. Check out a code sample of this project.
4) I created a new performance metric, named O-value, which measures predictive models' stability (work in process). Take a sneak peek here.
5) Contradict to current hype about deep learning, I have shown that recent adoption of deep learning in student modeling is neither superior in performance nor interpretable. Other researchers also reached the same conclusion independently.