froyosng/hiv-funding-optimisation-nsga2

Bi-objective NSGA-II optimisation of HIV intervention funding across 46 African countries (40.018 HST project, SUTD) # HIV Intervention Funding Allocation — Multi-Objective Optimisation with NSGA-II Project for 40.018 Heuristics and Systems Theory (SUTD, Summer 2026). We formulate the allocation of a fixed USD 200M budget across four HIV intervention types (ARV procurement, supply chain, workforce training, HIVDR surveillance) in 46 African countries as a bi-objective, non-convex nonlinear program: maximise aggregate DALYs averted while minimising cross-country inequity. The main method is NSGA-II, benchmarked against a greedy marginal-DALY heuristic, simulated annealing (weighted-sum scalarisation), and an optional reduced-instance Gurobi exact benchmark. Full methodology, results, and discussion are in `report/`. ## Key finding The scenario is structurally underfunded: total ARV need across the 46 countries is ~USD 872M against a USD 200M budget. Under that scarcity, funding concentrates on countries with the highest **marginal** DALY return per dollar — not the highest absolute burden. Nigeria (2.0M PLHIV) receives ~37% of the budget at the NSGA-II knee point, while South Africa (7.8M PLHIV, the largest burden in the dataset) receives close to nothing. This is a consequence of the saturating impact-response function, not a modelling error — see `report/` Sections 8 and 10 for the full mechanism and its ethical implications. ## Repository structure ``` hst_hiv_optimisation_full.py Main script: data loading, NSGA-II, greedy, simulated annealing, optional Gurobi benchmark, sensitivity analysis, all plots/CSVs requirements.txt Python dependencies hiv_data.xlsx UNAIDS AIDSinfo 2025 (2024 estimates) edited IHME-GBD_2023_DATA-*.xlsx IHME GBD 2023 HIV/AIDS DALY estimates (pre-split across the 4 interventions) outputs_hiv_optimisation/ Headline results: Pareto fronts, convergence plots, method comparison, sensitivity analysis summary and plots, per-country allocation CSVs report/ HST_final_report.pdf Compiled report (student IDs …