This IQP was an eclectic development of various features for the ASSISTments online learning platform. We designed features to identify student gaming behaviors, used trigram matching to determine similarity between hints and explanations of problems, concatenated data to find streaks of correctly answered problems, and clustered Common Core descriptions based on embeddings from MathBERT. We also simulated using deep Bayesian bandits to recommend content in the form of supports to struggling students. Our models were able to predict whether or not a student would get the next problem correct more frequently than random using an epsilon-greedy (RMS) model. All features were completed successfully and integrated into the ASSISTments Automatic Personalized Learning Service (APLS). These results all had significant findings to be expanded upon in further research.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Scopetski, Caleb
• Hendrickson, Matthew

Subject

• Project-Based Learning
• Inclusion
• Computing
• Innovation
• Learning

Publisher

• Worcester Polytechnic Institute

Identifier

• E-project-080922-132018
• 71676

Advisor

• Heffernan, Neil

Year

• 2022

UN Sustainable Development Goals

• 4 - Quality Education
• 10 - Reducing Inequality

Date created

• 2022-08-09

Resource type

• Interactive Qualifying Project

Source

• E-project-080922-132018

Rights statement

• In Copyright

Last modified

• 2022-12-20