My research focuses on machine learning and data mining for human-centered applications, such as educational and recommendation systems.
I develop models and algorithms to efficiently utilize variety and heterogeneity of data in these applications, while dealing with application-specific challenges, such as data uncertainty and sparsity. More specific research interests include:
Educational Data Mining:
* Knowledge Modeling and Performance Prediction: developing knowledge tracing models and algorithms, via tensor factorizations, deep sequential models, and point process modeling, specifically tailored to accurately and efficiently capture the sequential nature of learner knowledge gain and forgetting, estimate learning material’s domain knowledge map, and predict students’ future performance.
* Learner Behavior and Procrastination Modeling: creating discriminative factorization models and process models to distinguish efficient vs. inefficient learning behaviors in online learners, detect procrastination behaviors in learners, and study their effect on learning gain and performance.
* Personalized Learning: building educational recommender systems that rely on multi-aspect optimization methods to balance between depth and breadth of learning for students.
* Cross-Domain Recommendation: developing transfer learning and domain adaptation approaches to model user interests across multiple domains and systems for higher-quality suggestions.
* Recommendation Explanations: building models to explain recommendations to users, e.g., by generating reviews, and studying their effect on user perception of the recommendation systems.
* Augmenting Recommendation Systems with External Information: building algorithms that embed social connections, external user profiles, or semantic information to better capture user interests.