About This Project
Using AI to bring price transparency to Kenya's agricultural markets
Project Overview
The Kenya Agricultural Commodity Price Forecasting system is an AI-powered platform designed to predict retail and wholesale prices for 19 key agricultural commodities across 100+ markets in 30 counties.
Built using data from the Kenya Agricultural Market Information System (KAMIS), the platform combines three machine learning models — XGBoost, LSTM, and SARIMA — to deliver accurate short and medium-term price forecasts.
The goal is to empower farmers, traders, and consumers with actionable price intelligence that supports smarter buying, selling, and policy decisions.
41,690
Data Points
19
Commodities
100+
Markets
30
Counties
Data Source
KAMIS — Kenya Agricultural Market Information System
All price data is sourced from KAMIS, which collects weekly retail and wholesale prices from markets across Kenya. The dataset covers January 2024 to February 2026 and includes 41,690 price records aggregated monthly per commodity and market.
AI Models Used
XGBoost
Gradient boosting with lag features. Best for stable commodities with clear price patterns. Trained separately for retail and wholesale prices.
Best MAE: 3.32 (Beans Yellow-Green)
LSTM
Long Short-Term Memory neural network. Captures long-term seasonal dependencies. Especially strong on rice, maize flour, and dry onions.
Best MAE: 2.59 (Dry Onions)
SARIMA
Seasonal ARIMA baseline model. Handles trend and seasonality decomposition. Available for commodities with sufficient historical data.
Baseline statistical model
Meet the Team
Victor Kipkemboi
Project Lead
Jose` Barasa
Fullstack Developer
Sharon Gichira
Data Analyst
Graham Akello
ML Engineer
Maureen Kitonyi
ML Engineer
Charity Kanyua
ML Engineer
Explore Price Forecasts
Get AI-powered predictions for any commodity across Kenya.
Get Forecast →