This episode of Forecasting Impact features Professor Lauren B. Davis discussing her research on applying stochastic modeling and forecasting to food bank operations. Lauren shares how she began forecasting with a local food bank, which led her to focus on forecasting the highly uncertain supply of food donations. She details the food banks' donation sourcing process, the management of their supply chains, and the application of models like exponential smoothing, support vector regression, and ensemble methods to predict donation volumes.

Professor Davis addresses challenges in forecasting at various aggregation levels (network vs. location-specific), using optimization models for equitable allocation of limited supply, and the significance of storage and agency capacity limits. She emphasizes the importance of equity as an objective, the complexity of modeling true demand, and the crucial role of visual analytics and co-design with food bank partners. The episode underscores the practical impact of forecasting in humanitarian supply chains and the necessity of linking models with operational decisions.

In this episode, we sit down with Rami Krispin, a data scientist at Apple and active producer in forecasting, to explore his journey into forecasting and data science. He shares what first sparked his interest in the field and how that passion led him to develop key contributions, including the Hands-On Time Series Analysis with R book and the TSstudio package. We discuss his motivation for writing the book, who it’s for, and how TSstudio and other R packages he has developed have helped practitioners in the forecasting space. He also gives us a sneak peek into his upcoming book, Applied Time Series Analysis and Forecasting with R, and the new topics it will cover.

We then dive into the challenges of deploying forecasting models at scale and the role of MLOps in making machine learning projects production-ready. As a Docker Captain, our guest explains how Docker has changed his approach to time series forecasting and MLOps. We also discuss best practices for forecasting, common mistakes practitioners make, and strategies for improving reproducibility. Looking ahead, we talk about where time series forecasting is heading, the differences between R, Julia, and Python in this space, and how each ecosystem serves different needs.

You can follow his work on LinkedIn, subscribe to his newsletter, and stay updated on his latest projects.

Website: https://linktr.ee/ramikrispin
LinkedIn Page: https://www.linkedin.com/in/rami-krispin/

In this episode, Dr. Laila Akhlaghi and Professor Bahman Rostami-Tabar host a discussion on healthcare product forecasting in Eastern Africa with Harrison Mariki from Tanzania and Danielson Kennedy Onyango from Kenya. Harrison, founder of Afya Intelligence, discusses leveraging AI to improve forecasting for 7,000 primary healthcare facilities in Tanzania, addressing data quality and supply chain challenges. Ken, from inSupply Health, highlights the use of open-source tools and human-centered design to enhance forecasting accuracy and efficiency in Kenya. Both emphasize the importance of local talent, trust, and co-creation in developing effective forecasting solutions.