An Integrated Feedforward Neural Network for Categorical Prediction of Greenhouse Tomato Yield under Nigeria’s Climatic, Soil, and Agronomic Parameters

Abstract

Accurate prediction of tomato yield in greenhouse environments is essential for sustainable agriculture, particularly under Nigeria’s unique climatic, soil, and agronomic conditions. This study presents an integrated Feedforward Neural Network (FNN) model for the categorical prediction of greenhouse tomato yield, classified into low, medium, and high. The model integrates heterogeneous datasets encompassing climatic, soil, and agronomic features through a unified network architecture, data preprocessing, regularization, and cross-validation, which are employed to enhance generalization and predictive accuracy. The FNN, chosen for its simplicity and computational efficiency, achieved an overall accuracy of 93%, with strong precision, recall, and F1-scores across yield categories. These results highlight the potential of the proposed model for data-driven yield prediction and sustainable greenhouse management in Nigeria.