Perbandingan Optimasi Algoritma Support Vector Machine dengan Metode Bagging dan Adaboost untuk Klasifikasi Kematangan Buah
Abstract
Dependence on manual methods for determining the ripeness of fruits like Siam oranges often leads to issues of subjectivity, inefficiency, and potential human error. Addressing this, this research was developed by creating an automatic ripeness classification system using the Support Vector Machine (SVM) algorithm, supplemented by ensemble learning approaches with Bagging and Adaboost. This approach aims to enhance the model's performance to be more robust against variations in lighting conditions. The focus of this research is to determine the performance and identify the best hyperparameter combination. The dataset used is a primary dataset collected from a Siam orange orchard in Banua Village, Kintamani, Bali. Consisting of 354 images, the data was divided into two parts: 80% for training and 20% for testing, undergoing a color feature extraction process for each color channel (RGB), including mean, variance, and mode. Experimental results show that the SVM + Adaboost model (kernel=linear, c=100, num_learners=30) achieved the highest classification performance with an accuracy of 94.4%, precision of 94.4%, recall of 94.2%, and an F1-score of 94.16%, outperforming the SVM + Bagging model (93% accuracy) and the single SVM model (91.5% accuracy). This indicates that the addition of optimization methods can improve model performance in the case of Siam orange ripeness classification
