PlanetTerp Predictor
An AI-powered tool that forecasts professor grades and course difficulty using historical data from PlanetTerp, helping students make informed registration decisions.
Overview
The PlanetTerp Predictor is an NLP-driven data science project that analyzes student sentiment to predict professor ratings. By processing thousands of reviews from the PlanetTerp API, the tool identifies key indicators of teaching quality and grade distributions.
Highlights
Built an NLP pipeline predicting professor ratings using 4,544+ reviews, achieving a 0.788 R-squared accuracy score.
Engineered features from large-scale UMD course data; optimized Random Forest and Gradient Boosting hyperparameters.
Implemented text processing to extract sentiment scores and correlate them with quantitative rating metrics.
Results
The final model demonstrated that student sentiment in textual reviews is a strong predictor of numerical ratings, with Gradient Boosting providing the best performance across various course disciplines.