Optimizing Agricultural Management Practices for Maize Crops: Integrating Clusterwise Linear Regression with an Adaptation of the Grey Wolf Optimizer

Effectively managing agricultural practices is crucial for maximizing yield, reducing investment costs, preserving soil health, ensuring sustainability, and mitigating environmental impact. This study proposes an adaptation of the Grey Wolf Optimizer (GWO) metaheuristic to operate under specific constraints, with the goal of identifying optimal agricultural practices that boost maize crop yields and enhance economic profitability for each farm. To achieve this objective, we employ a probabilistic algorithm that constructs a model based on Clusterwise Linear Regression (CLR) as the primary method for predicting crop yield. This model considers several factors, including climate, soil conditions, and agricultural practices, which can vary depending on the specific location of the crop. We compare the performance of the Grey Wolf Optimizer (GWO) algorithm with other optimization techniques, including Hill Climbing (HC) and Simulated Annealing (SA). This analysis utilizes a dataset of maize crops from the Department of Córdoba in Colombia, where agricultural practices were optimized. The results indicate that the probabilistic algorithm defines a two-group CLR model as the best approach for predicting maize yield, achieving a 5% higher fit compared to other machine learning algorithms. Furthermore, the Grey Wolf Optimizer (GWO) metaheuristic achieved the best optimization performance, recommending agricultural practices that increased farm yield and profitability by 50% relative to the original practices. Overall, these findings demonstrate that the proposed algorithm can recommend optimal practices that are both technically feasible and economically viable for implementation and replication.