Researchers in EIU's Faculty of Computing & Information Technology have published a machine-learning model that forecasts maize and teff yields for smallholder farms using nothing more than satellite imagery and a smartphone weather log. The study — a collaboration with agricultural extension officers in Oromia — was released through the EIU Research Portal after passing the university's dual-approval workflow (faculty head and academic board).
The gradient-boosted ensemble achieved 91% accuracy in back-testing across three growing seasons, and the team has released the dataset openly. "Our goal was a tool a farmer's cooperative can run without a data scientist," said lead author Dr. Samuel Kefale. "It forecasts yields 60 days before harvest, early enough for price negotiation and food-security planning."
The full paper — IMRaD structure, datasets and methodology — is downloadable from the EIU Research Portal, and the team invites replication studies from partner universities.
