The data consists of historical wholesale and retail prices of maize at different markets in Kenya from 2021 - 2025. The first dataset is from KAMIS. It consists of historical prices for three types of maize (white, yellow and mixed-traditional). You can pull extra data from the KAMIS website to complement the sample that is provided. The second dataset, collected by agriBORA, is the transaction data between businesses showing the wholesale price of white maize for a given week.

Using historical prices of dry maize in Kenya, your task is to develop a machine learning solution to predict average weekly prices of maize in the counties of Kiambu, Kirinyaga, Mombasa, Nairobi and Uasin-Gishu. agriBORA data is the data that needs to be forecasted, however, KAMIS data is provided if you would like to supplement the agriBORA data. At each prediction step, your model should generate forecasts for two consecutive weeks. The forecasting period spans six consecutive weeks, from November 17, 2025 to January 10, 2026.

You are encouraged to explore and incorporate any external data you believe may improve your solution - precipitation, vegetation indices (NDVI), or other relevant data could provide useful signals. Part of the challenge is discovering which features truly help predict prices. Any additional dataset used must be open-source, and must be shared in the chat in order for us to consider your winning submission.

If you choose to use additional data sources, please submit the following:

• Scripts for downloading or pulling the data
• The raw data itself
• Any processing scripts you applied
• The processed data used as input to your model

This ensures your solution is fully reproducible and allows others to understand and build upon your approach.