Selected research

01 — 03

01 / Healthcare Operations

Chicago Booth · UChicago Medicine

Improving access to outpatient care

Working paper

Estimating Provider Episodes as Operational Units of Outpatient Access

Duygu Soylemez, Dan Adelman

In preparation

How can clinics make room for new patients while maintaining care for returning patients? I analyze outpatient visit histories and develop capacity-planning models to support that decision.

My work involves processing 4.8 million outpatient encounters to construct 1.27 million provider episodes, using more than 80 source datasets totaling 250 GB. These episodes help characterize patient visit patterns, cancellations, and continuity of care.

  • Large-scale data analysis
  • Survival analysis
  • Optimization
  • Simulation

Presentations

  • INFORMS Annual Meeting (2025) — Atlanta, USA
  • INFORMS Healthcare Conference (2026) — Raleigh, USA
Approach & research context

I developed a method combining survival analysis and importance sampling to estimate episode-duration distributions from incomplete patient histories, and a data-driven optimization model for outpatient capacity planning.

I am also building a simulation model to evaluate an alternative scheduling system intended to improve new-patient throughput and waiting times while preserving service for returning patients. The proposed system is being studied ahead of a pilot.

Advised by Dan Adelman.

02 / Optimization & Algorithms

Chicago Booth

Network design for online retailing

Working paper · 2026

Local and Global (Approximate) Submodularity in Network Design

Levi DeValve, Duygu Soylemez

Submitted to Management Science

Large fulfillment networks create difficult decisions about where to place items and how to serve orders. I study the mathematical structure of these problems to develop efficient approximation algorithms.

My research uses linear-programming-based surrogate functions and local search, exploiting local and approximate submodularity to obtain constant-factor approximation guarantees.

  • Approximation algorithms
  • Linear programming
  • Network optimization

Presentations

  • POMS Annual Meeting (2023) — Minneapolis, USA
  • INFORMS Annual Meeting (2024) — Seattle, USA
Approach & research context

I formulate network-design problems as structured set-optimization models with bipartite coverage and fulfillment-flow constraints. This work develops a constant approximation framework for multi-item order fulfillment.

Advised by Levi DeValve.

03 / E-commerce Operations

Bilkent University

Assortment planning with fewer split orders

Master's thesis · 2021

Assortment Planning Considering Split Orders

Duygu Soylemez

M.S. in Industrial Engineering, Bilkent University

Read thesis (PDF)

When an order is split across shipments, fulfillment can become more costly and delivery more complex. I developed assortment-planning optimization models and heuristics that aim to reduce split shipments.

In this work, I characterized the structure of the linear-programming relaxation and developed an algorithm to find its optimal solution.

  • Assortment planning
  • Mathematical programming
  • Heuristics

Presentations

  • National OR/IE Congress (2019) — Ankara, Turkey
  • National OR/IE Congress (2021) — Online (Boğaziçi University, Istanbul, Turkey)
  • INFORMS Annual Meeting (2022) — Indianapolis, USA
Approach & research context

This research began during my undergraduate studies and continued through my master's at Bilkent University, connecting optimization theory to order-fulfillment decisions.

Advisor: Oya Karasan. Co-advisor: Alper Sen.