A statistical analysis and predictive modeling project in R, examining how lifestyle and clinical factors relate to diabetes risk using real public health survey data. Combines formal hypothesis ...

We employed a multi-step feature selection strategy combining L1-regularized logistic regression with Recursive Feature Elimination (RFE) to identify key stable features associated with adverse ...

Binary sentiment classifier built on the IMDB Movie Reviews dataset. Ships with a trained Logistic Regression, Naive Bayes, and Linear SVM — plus a Streamlit web app, a FastAPI backend, and a Jupyter ...

Developed in Python with LangGraph and Streamlit, the system translates user questions into optimized SQL queries, validates them with dry-run checks, enforces guardrails such as partition filters, ...

The architecture offers a tiered approach with three distinct access levels: "Basic" mode is designed for users with SQL skills who prioritize rapid data discovery through Dremio. "Expert" mode ...

Our three-level multinomial logistic framework, trained separately for each possible outcome and combined through a principled normalization procedure, was evaluated on held-out data. The framework ...

aInstitute of Computer-Assisted Cardiovascular Medicine, Deutsches Herzzentrum der Charité Augustenburger Platz 1, 13353 Berlin, Germany bDepartment of Congenital Heart Disease – Pediatric Cardiology, ...