A new hybrid model increasing productivity in agriculture using deep learning method

Today, the production of agriculture products and ensuring its continuity are of critical importance. Additionally, the productivity of the products during the production stage is highly significant. A high yield of the product will not only reduce the farmer's product and financial loses but also provide quality products to the consumers. With the advancements in technology in recent years, studies have been conducted utilizing machine learning and deep learning models to determine the productivity of agricultural products. In this study, a new hybrid model based on deep learning was designed to assess the productivity of agricultural products based on their quality. The designed hybrid model involves extracting feature images using CNN, a deep learning model, and classifying them with machine learning methods. Consequently, the optimal hybrid model was determined. For this study, two different datasets were used: Elma Veri Kümesi and FruitsGB. Elma Veri Kümesi is divided into four different dataset scenarios (Dataset_A, Dataset_B, Dataset_C, Dataset_D), and the FruitsGB Dataset is divided into four different dataset scenarios (Meyve_A, Meyve_B, Meyve_C, Meyve_D). Accuracy, precision, sensitivity, specificity, F1 score, balanced accuracy and Kappa scores and ROC analysis were performed for all dataset scenarios. The designed hybrid model, which achieved the highest accuracy rate of 99.80%, was obtained in the Dataset_C scenario using CNN-SVM (linear) model in Elma Veri Kümesi. The designed hybrid model, which achieved the highest accuracy rate of 99.30%, was obtained in the Meyve_C dataset using CNN-SVM (rbf) model in FruitsGB Dataset. Considering the results of the study, the hybrid model exhibited effective outcomes.