Evaluation of optimization techniques with support vector machine for identification of dry beans

Abstract

Automated recognition or categorization of fruits or crops through image processing technology is a highly sought-after technique in agriculture.
This approach has been demonstrated to outperform manual monitoring in humans. It reduces the time and expenses incurred by farmers and ensures that end users receive high-quality products. This study details the process for identifying dry beans. We employed a support vector machine (SVM) model and three optimization techniques: Bayesian optimization, random search optimization, and grid search optimization. This study aimed to determine the optimal optimization technique for identifying dry beans using an SVM model. Various metrics, including the validation accuracy, test accuracy, total validation cost, total test cost, and minimum classification error, were computed and analyzed to compare the performance of each optimization method. In comparison, Bayesian optimization was determined to be the most effective optimization technique for identifying dry beans, with a validation accuracy of 93.06%, test accuracy of 92.29%, and minimum classification error of 0.069384.