Development of a Predictive Model for Aflatoxin Contamination Risk in Groundnut Supply Chains in Kano and Kaduna States, North West Nigeria

Aflatoxin contamination of groundnuts poses a major food safety, public health, and trade competitiveness challenge in Nigeria, where groundnut production is concentrated in North West and North Central states. Predictive modelling of contamination risk across supply chain nodes offers a scientific basis for targeted intervention. This study developed and validated a predictive model for aflatoxin contamination risk in groundnut supply chains in Kano and Kaduna States, North West Nigeria. A longitudinal supply chain tracking design was employed, with 200 groundnut consignments traced from farm to retail over two agricultural seasons. At each supply chain node, aflatoxin content was quantified using high-performance liquid chromatography, and environmental and handling parameters including temperature, relative humidity, moisture content, insect damage, and storage duration were recorded. Predictive models were constructed using machine learning algorithms including random forest, gradient boosting, and artificial neural networks, and compared against a classical regression baseline. Model performance was evaluated using area under the receiver operating characteristic curve, sensitivity, specificity, and calibration statistics. The gradient boosting model achieved the highest predictive accuracy with AUC of 0.89. Key predictors included post-harvest moisture content, storage temperature above 28 degrees Celsius, storage duration beyond three weeks, and visible insect damage at aggregation nodes. The validated model provides a practical decision-support tool for aflatoxin risk management in the Nigerian groundnut sector. Keywords: aflatoxin, groundnut supply chain, predictive model, machine learning, North West Nigeria