Maize cultivation in North-East Nigeria experiences difficulties due to ineffective management of maize farm inputs such as labour, herbicides, fertilizer, pesticides, land rent and seeds which results to poor yields. This affects food availability as well as farmers’ income. Random Forest (RF), a machine learning (ML) technique, is applied in this study to manage the inputs and maximize production. Primary data on the inputs and maize yield were obtained and analysed. The methodology followed Cross Industry Standard Process for Data Mining (CRISP-DM), including data pre-processing, feature selection, model training, and validation. The RF model was trained on primary data and compared with Ordinary Least Square Regression (OLSR) method to predict optimal input combinations for maximum yield. The results revealed that an increase of less than one unit in input can double the output. Correlation exploration showed that labour (0.23), herbicides (0.17) and fertilizer (0.12) had the maximum impact on output. The study recommended that, data-driven digital agriculture should be practiced and there should be increase in fertilizer and herbicide application to realize viable productivity. By applying machine learning for yield forecast and optimization, farmers can enhance efficiency, optimize profitability, and improve food security. The result of the evaluation metrics - Mean Square Error (MSE) and Mean absolute Error (MAE) revealed that RF is more reliable for digital agriculture than OLSR Model. Subsequently, RF should be applied to guide farmers for data-driven decision-making in ideal input application and yield maximization.