Analysis of Agricultural Yield Prediction Using Regression Models in Kano State

Improving food security in northern Nigeria requires reliable predictions of crop yields that can guide policy and farm management decisions. This study analysed the effectiveness of regression models for predicting agricultural yields in Kano State, North West Nigeria, using a combination of climatic, soil, and agronomic variables. Secondary data covering ten years of sorghum and maize production statistics were obtained from the Kano State Agricultural Development Authority and the Nigeria Meteorological Agency for annual rainfall, temperature, and humidity records. Multiple linear regression, polynomial regression, and ridge regression models were fitted and compared using Root Mean Square Error and R-squared metrics. The polynomial regression model achieved the highest predictive accuracy for maize yield, with an R-squared value of 0.81 and an RMSE of 0.34 tonnes per hectare. Annual rainfall and soil nitrogen content were the most statistically significant predictors across all models. Ridge regression was particularly robust in handling multicollinearity between climatic variables. The study identified that delayed onset of rains in the preceding three years was strongly associated with yield decline exceeding 25 percent. Data quality inconsistencies in manually recorded production reports were identified as a key limitation. The findings have direct implications for early warning systems and agricultural advisory services in North West Nigeria. Recommendations include investing in automated meteorological data collection infrastructure, building regression-based decision support tools for extension officers, and establishing a centralised state-level crop yield database to support future predictive modelling efforts.

Keywords: regression models, agricultural yield prediction, Kano State, food security, predictive analytics