Miss Udeshi Salgado

(she/her)

Research

Thesis

Global immunisation supply chains: integrating hybrid AI models into inventory optimisation considering non-monetary welfare metrics

Background: Approximately 500,000 children in developing countries die annually because of limited access to immunisations (Gavi, 2022). This devastating reality is further intensified by numerous factors, such as inaccurate forecasts, stock-outs, and missed opportunities. The traditional forecasting and inventory methodologies is characterized by a one-size-fits-all approach and face several challenges, including inadequate or inaccurate data and the reliance on unrealistic assumptions on population growth. Additionally, the forecasts are only generated at a national level, limited to yearly projections, and not integrated into inventory optimisation. Hence, their usefulness is constrained. Therefore, Immunisation programs are exploring alternatives to improve forecasting and inventory processes and enhance supply chain efficiency.

Suggested aim: We propose a Hybrid AI approach to forecast vaccine needs that combines domain-specific knowledge with probabilistic forecasts using Quantile Regression Averaging, which produces accurate probabilistic forecasts for the entire supply chain hierarchy. Forecasts then will be integrated into the inventory optimisation that accounts for non-monetary well-being metrics in the loss function, in addition to monetary costs. The project will be empirically routed in and validated using available and forthcoming data of immunization programs from an African country. The solution may then be adopted in other developing countries given similar immunization programs. There is scope for the student to further develop and refine the research aim and design, which may include interviews with immunisation providers to gain information for research design.

Funding sources

Engineering and Physical Sciences Research Council (EPSRC)