
Certificate of Completion
THIS ACKNOWLEDGES THAT
HAS COMPLETED THE FALL 2026 DATA SCIENCE BOOT CAMP
Kevin Dao
Roman Holowinsky, PhD
November 15, 2026
DIRECTOR
DATE

TEAM
Predicting Hospital Readmissions for Diabetic Patients
Kevin Dao, Taylor Tan

This project predicts readmission within 30 days for patients hospitalized with diabetes. We explored logistic regression, CatBoost, random forests, and XGBoost, then developed a compact logistic regression using seven encounter variables, including a primary diagnosis grouped from ICD-9 codes. We kept repeat encounters from the same patient together across data splits and evaluated models using F2 score, which emphasizes recall. On the test set, the seven-input model scored 0.42477, compared with 0.41277 for a 47-predictor baseline. It produced fewer false positives while identifying slightly fewer readmissions. Its short input list offers a simpler model for further study.
