2253. Enhancing Routine Immunisation Demand Forecasting: A Hybrid Probabilistic and Domain-Specific Approach
Invited abstract in session TA-38: Forecasting, prediction and optimization 1, stream Data Science meets Optimization.
Tuesday, 8:30-10:00Room: Michael Sadler LG19
Authors (first author is the speaker)
| 1. | Udeshi Salgado
|
| Cardiff Business School, Cardiff University |
Abstract
Accurate demand forecasting is crucial for effective routine immunisation supply chains. In low- and middle-income countries (LMICs), traditional methods often rely on outdated population estimates, static wastage assumptions, and simplistic stock adjustment heuristics. These methodological limitations contribute to stock imbalances, missed vaccination opportunities, and unreliable supply projections, ultimately reducing immunisation coverage and increasing the prevalence of vaccine-preventable diseases. We propose a hybrid intelligence framework integrating domain-specific knowledge with probabilistic forecasting and machine learning to improve vaccine demand prediction. Our approach replaces static population estimates with dynamic survival birth rate modeling and incorporates consumption-based forecasting to capture missed vaccination opportunities. This ensures a more adaptive and data-driven supply chain by quantifying uncertainty, enabling proactive inventory adjustments, and reducing missed vaccination opportunities. We evaluate our model through simulations and real-world routine immunisation supply chain data. This framework holds promise for enhancing forecasting accuracy, enabling more confident supply planning, and ultimately improving routine immunisation coverage.
Keywords
- Forecasting
- Machine Learning
- Supply Chain Management
Status: accepted
Back to the list of papers