Publications, revisions, and current work

European Journal of Operational Research, 328(1), 282-294

We propose a cost-sensitive ensemble of logistic regressions that minimizes expected decision cost while retaining the interpretability of logistic regression coefficients.

Why it matters. It shows that better decision performance does not have to come from abandoning interpretability; the gain can come from diversity-aware ensemble design under a cost-sensitive objective.

Under revision at Information Sciences (JCR Q1 / CAS Q1 Top)

This work extends ensemble cost-sensitive logistic regression to real-world datasets with nominal, ordinal, and continuous predictors through a mixed-type regularization penalty.

Why it matters. It connects interpretability with structured feature handling, making cost-sensitive modeling more usable on practical mixed-type business data.

Work in progress

This project introduces sparse regularization into prescriptive analytics to select a small set of representative prototypes from historical data.

Why it matters. It aims to make data-driven prescriptions faster, more interpretable, and closer to how people justify decisions through representative cases.