MLP–SVM Hybrid Framework for Agricultural Commodity Price Prediction

Accurate agricultural commodity price prediction remains a challenging task due to nonlinear market dynamics, regional variability, and frequent price fluctuations. While traditional machine learning models provide reasonable performance, they often struggle to balance feature representation and classification robustness. Building upon the optimal feature subset obtained through hybrid metaheuristic feature selection in earlier phases, this study proposes a novel MLP–SVM hybrid classification framework for agricultural commodity price prediction. In the proposed model, a Multi-Layer Perceptron (MLP) is employed to learn deep and nonlinear feature representations from preprocessed commodity price data, while a Support Vector Machine (SVM) is utilized as the final classifier to ensure robust decision boundaries and improved generalization. The hybrid architecture effectively combines the representation learning capability of neural networks with the margin-based strength of SVMs. Experiments conducted on real-world agricultural commodity price datasets demonstrate that the proposed MLP–SVM model outperforms individual classifiers in terms of accuracy, precision, recall, and F1-score. The results confirm that integrating deep feature learning with classical machine learning classifiers significantly enhances price-range classification performance, offering a reliable and scalable solution for agricultural price forecasting systems.