Designing Optimal Dynamic Treatment Regimes Using TMLE for Personalized Math Course-Taking Plans

• In Event: AERA Roundtable Session Friday 9:50 am Four Seasons Ballroom 4
  In Roundtable Session: Machine Learning Techniques I: Advancing Equity, Prediction, and Measurement (Table 4)

Abstract

This study provides an approach to designing optimal dynamic treatment regimes (DTRs) using Targeted Maximum Likelihood Estimation (TMLE), coupled with ensemble learning algorithms, to build a personalized recommendation model for high school math course-taking plans. Our method uses backward induction and feasibility constraints to create personalized, data-driven recommendations under practical considerations. Our simulation study demonstrates that the proposed DTR-TMLE method yields more accurate recommendations compared to Q-learning based on linear regression. We apply the TMLE method to design math course recommendations using data from the High School Longitudinal Study of 2009 (HSLS:09), ultimately aiming to recommend the right math course for each student at the right time.

Authors

• Chenguang Pan, Teachers College, Columbia University

• Youmi Suk, Teachers College, Columbia University