Developer · 2024
Healthcare Inventory MDP (Knowledge Representation & Reasoning)
Formal Markov Decision Process model for hospital drug inventory management. Applies value iteration and policy iteration to derive optimal ordering policies under demand uncertainty, with simulation-based evaluation.
Overview
Formal Markov Decision Process (MDP) model for hospital drug inventory management. The system applies value iteration and policy iteration to derive optimal ordering policies under demand uncertainty, with simulation-based evaluation against baseline heuristics.
The question
Hospitals must balance drug availability against holding and shortage costs under stochastic demand. Manual heuristics are common but sub-optimal.
My approach
Built a finite-state MDP with explicit transition and reward matrices, implemented dynamic programming solvers (Value Iteration and Policy Iteration), and compared optimal policies against baseline heuristics using Monte-Carlo simulation.
Outcome and limits
Reported simulation result: approximately 8.6% lower mean cost than a baseline heuristic across 50 runs (2280.8 versus 2495.6).
A simulation comparison under project assumptions; not evidence of savings in a deployed healthcare service.
Interested in this work? Get in touch.