Modeling Student Knowledge over time - Riiid! Data Science Competition

Used an extremely simple Elo Rating algorithm to model student knowledge on an e-learning platform, which achieved higher scores than the baseline of state-of-the-art self-attentive knowledge tracing or gradient boosting algorithms (SAKT, SAINT+, LightGBM), while having significantly higher interpretability, and being much more adapted to a potential real-time production deployment.