A Hybrid Voting Ensemble Model for the Efficient Sorting and Classification of Date Fruit Varieties Un modelo híbrido de votación por conjunto para la clasificación y ordenamiento eficiente de variedades de dátiles

Dates constitute one of Algeria's most important agricultural products, given their substantial health and economic advantages. Furthermore, they represent a vital export item beyond the hydrocarbon industry. The existing conventional techniques for classifying and sorting dates are ineffective, time-consuming, and labor-intensive, leading to a discrepancy between restricted exports and elevated production levels. This work presents an ensemble learning (EL) model that utilizes transfer learning (TL) strategies to overcome obstacles and improve date fruit classification. We assess the efficacy of four classifiers, i.e., MobileNetV2, EfficientNet, DenseNet201, and an ensemble soft voting classifier that employs TL, with a dataset of 1619 photos representing 20 distinct varieties of Algerian dates. The dataset used is one of the greatest benchmarks for varietal diversity. The suggested hybrid model exhibits exceptional performance, with a validation accuracy of 99.07% and a classification accuracy of 99.93%. It establishes a new benchmark in agricultural technology by exceeding all assessed models in terms of precision, recall, and F1-score. These findings demonstrate the potential of this approach to completely transform date sorting and markedly improve agricultural production and efficiency.