Improving the Performance of a Support Vector Machine by Using Features Extracted from MobileNetV2 for Identifying Coffee Plant Diseases in Pagar Alam

Nurmaleni Nurmaleni (Informatics Engineering, Institute of Technology Pagar Alam)
Febriansyah Febriansyah (Informatics Engineering, Institute of Technology Pagar Alam)
Inka Rizky Padya (Agricultural Product Technology, Institute of Technology Pagar Alam)
Desi Puspita (Informatics Engineering, Institute of Technology Pagar Alam)

Bulletin of Network Engineer and Informatics (BUFNETS)
Vol. 4 No. 2 (2026): BUFNETS (Bulletin of Network Engineer and Informatics) October 2026

Delays and inaccuracies in the early detection of plant diseases contribute to the decline in coffee productivity in Pagar Alam City. This study aims to optimize the performance of the Support Vector Machine (SVM) by utilizing MobileNetV2 as a feature extractor to improve the accuracy of coffee leaf disease classification. The research method follows the CRISP-DM framework, which includes problem understanding, data preparation, image preprocessing, data augmentation, feature extraction using MobileNetV2, classification using SVM, and model evaluation. The dataset consists of three classes: healthy leaves, leaf spots, and leaf rust, with all images resized to 224×224 pixels before processing. Test results on 180 test data points showed that the SVM model approach using MobileNetV2 was able to achieve an accuracy of 95%, with a precision of 95%, a recall of 94%, and an F1-score of 94%—higher than that of conventional SVM, which achieved an accuracy of 84%. The research results indicate that feature extraction using MobileNetV2 enhances the SVM’s ability to distinguish the visual characteristics of coffee leaf diseases more accurately and efficiently. This approach has the potential to be implemented as a fast, lightweight, and real-time coffee plant disease detection system to support increased coffee productivity in Pagar Alam City.