Machine learning and response surface methodology for optimizing olive waste compost in sustainable chickpea production

Purpose This study aims to combine machine learning and response surface methodology (RSM) for the optimization and prediction of the characteristics of compost made from olive mill waste cake (OMWC) residues. Design/methodology/approach After maturation, compost produced from OMWC residues was applied to chickpea cultivation. The plants, irrigated with rainwater, were monitored for vegetative growth, phenological stages, and production parameters. A variety of modeling techniques – including RSM, artificial neural networks (ANN), support vector machines (SVMs) and XGBoost – were used to identify the optimal compost application conditions for maximizing chickpea yield. Findings The results showed excellent performance of the optimization model with the Box-Behnken Design (RSM-BBD) and ANN, with a high R2 (0.9205 and 0.9718, respectively) value and a lower root mean squared error (8.0368 and 4.2833, respectively), while the SVM and the XGBoost models recorded lower values for the comparison elements used for both of the elements. Originality/value The study demonstrates that the choice of modeling method depends upon the specific nature of the problem and the required accuracy. To conclude, this study contributes to a better understanding of crop yield processes and paves the way for further optimization of OMWC residue compost performance for yield enhancement, with significant practical implications in agriculture.