Can we predict vegetable prices in Nepal three months ahead?

Researchers built a composite price index from 135 daily vegetable commodities in Kathmandu (2013–2023) and tested 14 forecasting models. Tree-based ensemble methods outperformed complex deep learning models, with the winning approach achieving 0.68% error on 90-day forecasts. For a volatile emerging market, this level of predictability could help governments and traders anticipate supply crises and stabilize food prices before they spiral.