Integration Of Machine Learning And IOT Smart Traps For Fruit-Fly Monitoring In Gerga Citrus Orchards Of Pagar Alam

Fruit-fly infestation is an important constraint in Gerga citrus production because it can reduce fruit quality, yield, and economic value. Conventional monitoring based on manual trap inspection is labor-intensive and provides limited real-time information for pest-management decisions. This study proposes an Internet of Things (IoT) Smart Trap integrated with machine learning for automated fruit-fly detection, counting, and population monitoring in Gerga citrus orchards in PagarAlam, Indonesia. The proposed system integrates image acquisition, environmental sensing, edge computing, YOLO-based object detection, EfficientNet-B0 verification, DeepSORT-based tracking, cloud data transmission, and XGBoost-based population prediction. The research follows a CRISP-DM-based workflow covering data acquisition, preparation, modeling, evaluation, and deployment. The evaluation showed that the proposed YOLO–EfficientNet-B0 framework achieved a precision of 97.1%, recall of 95.6%, F1-score of 96.3%, mAP@50 of 97.8%, and mAP@50–95 of 81.4%. Thetracking evaluation achieved a counting accuracy of 95.2%, while the XGBoost model obtained an MAE of 3.21, RMSE of 4.56, and R² of 0.89. These results indicate the potential of combining automated image detection, environmental sensing, tracking, and population prediction within an IoT Smart Trap for continuous fruit-fly monitoring. The proposed framework can provide a foundation for early-warning systems and support more precise Integrated Pest Management (IPM) decisions in Gerga citrus orchards.