Context:

Pulses have been considered an environmentally-friendly and inexpensive alternative to commodities like meat, ideal in helping to combat food insecurity. The FFPI (FAO Food Price Index— measure of international food commodity prices) suggests that effects of the COVID-19 pandemic have caused dramatic price surges in all food groups, including pulses, that threaten our ability to feed the growing global population. But is it really just the impacts of the pandemic that threatens this ability?

In order to answer the ultimate question of: “How did observed pulse prices differ from pre-pandemic price dynamics during COVID-19?”, I wanted to focus on the price forecasts of this food group in countries most susceptible to food insecurity as determined by the World Food Programme (WFP). I examine food price data from the WFP to create time series forecasts via gradient boosting of what could have happened to the prices of select pulses in the developing world if the pandemic had not occurred.

<aside> 🫘

Key Takeaways:

Forecasts predicted overall lower pricing for Pulses, suggesting the Pandemic triggered a series of unobserved events resulting in actual prices rising more than expected
Cyclical behaviors occurring in both forecasts and actual pricing suggest that there are deeper long term issues that need to be addressed in order to maintain viability of pulses as a food source. </aside>

Methods:

Multi-Step Ahead Forecasting with Multiple Time Series
Nested Cross Validation

Model:

Gradient Boosted Tree Model

Tools/ Packages Used:

Google BigQuery
Google Data Studio (2021 EDA Data Visualizations)
R Studio 4.1.0 (Forecasts & Forecast Visualizations)
- ggplot2 lubridate tidyverse forecastML dtplyr glmnet xgboost
Python 3.13.3 (2025 EDA Update)
- pandas numpy pygwalker
Kanaries AI
- graphic walker

Code: