SISTEM KLASIFIKASI PENYAKIT PADA DAUN LABU SIAM BERBASIS CNN DENGAN OPTIMASI HYPERPARAMETER DI DESA PUDAK KULON

Indonesia, as an agrarian country, relies heavily on the agricultural sector as the backbone of the community’s economy, particularly in mountainous regions such as Pudak Kulon Village. In this area, chayote has become one of the alternative commodities developed by the community after the livestock sector was affected by the foot-and-mouth disease (FMD) outbreak. However, chayote cultivation faces challenges from leaf diseases such as downy mildew and powdery mildew, which can significantly reduce both quality and yield. Manual disease detection has proven inefficient, error-prone, and time-consuming, requiring specific expertise. Therefore, this study aims to develop an automated and computerized leaf disease classification system for chayote using the Convolutional Neural Network (CNN) method with hyperparameter optimization. The model was trained on a dataset of leaf images consisting of three classes: downy mildew, powdery mildew, and healthy leaves. Performance evaluation using classification metrics showed excellent results, with an F1-score of 0.99 for downy mildew, 1.00 for powdery mildew, and 0.99 for healthy leaves. Both macro and weighted averages achieved 0.99, indicating high consistency across all classes. The results of this study produced a classification system that can assist farmers in early disease diagnosis, thereby improving crop productivity and reducing potential losses caused by chayote leaf diseases.