Deteksi Kecurangan pada CBT Online Menggunakan YOLOv8

Abstract

The current written exam is being administered differently due to advancements in technology. Computers are being used to administer written exams, also referred to as computer-based tests (CBT). E-learning is typically used for CBT exams, and an internet connection is required. Participants in online CBT have the flexibility to take the test at any time and from any location. Currently, there is inadequate supervision for each participant in online CBT. As a result, a system that can identify participant dishonesty is required. This research implemented the You Only Look Once (YOLO) version 8 algorithm to detect three types of fraudulent objects in online tests: phone, Other people and book with the amount of data used as much as 4119 image data and 4815 annotations performed. The annotation consists of 1882 data as data phone, 1563 data from other people and 1370 data books. According to the findings, the YOLOv8 algorithm has a 97% accuracy rate, 97.5% precision rate, 98.7% recall rate, and 98.1% F1-Score when it comes to real-time cheating object detection. As long as the object is in front of the camera at the ideal distance, this model may also identify many cheating objects in a single frame. These findings demonstrate how effectively the YOLO V-8 algorithm-based system has performed in identifying deceptive objects during online cognitive behavioral therapy.