Skripsi
PEMANTAUAN POLA PERILAKU LALU LINTAS BERBASIS JARINGAN CCTV BERDASARKAN TEKNOLOGI ANALISIS DATA MENGGUNAKAN YOLOV8 DENGAN ALGORITMA HIERARCHICAL CLUSTERING
This study presents the design of a traffic behavior monitoring system based on CCTV networks by comparing two algorithmic approaches: YOLOv8 for real-time object detection and Hierarchical Clustering for unsupervised object grouping without prior labeling. The system is developed to automatically identify behavioral patterns in traffic, such as violations involving riders not wearing helmets. Image and video data were collected from several strategic locations in Palembang and processed through stages of annotation, model training, and performance evaluation. The YOLOv8 model demonstrated high accuracy based on evaluation metrics including confusion matrix, precision, recall, and F1-score. Meanwhile, the Hierarchical Clustering results were visualized through dendrograms, serving as a comparative reference to the YOLOv8-based classification. The findings indicate that integrating both supervised and unsupervised methods offers complementary insights and can serve as a foundation for developing more robust, data-driven traffic surveillance systems.