Article Per Year (5 Year)
p-Index From 2021 - 2026
0.23
P-Index
This Author published in this journals
All Journal JSAI (Journal Scientific and Applied Informatics)
Uus Rusmawan
Universitas Dian Nusantara, Jakarta, Indonesia
Computer Science & IT

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
Model Deteksi Kecurangan Ujian Menggunakan Pendekatan SURF-CNN Berdasarkan Data Citra Digital Uus Rusmawan; Imam Mulya; Muchamad Sandy; Abd Rahman; Pupu Ramadhan
JSAI (Journal Scientific and Applied Informatics) Vol 8 No 3 (2025): November
Publisher : Fakultas Teknik Universitas Muhammadiyah Bengkulu

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.36085/jsai.v8i3.9383

Abstract

Detecting cheating during examinations was one of the main challenges in maintaining academic integrity in educational environments. This study developed an approach to detect cheating behavior by utilizing a combination of Speeded Up Robust Features (SURF) and Convolutional Neural Networks (CNN) based on digital images. The novelty of this research lay in the application of the SURF method as a feature extraction technique to detect suspicious objects and movements, which were then further analyzed using CNN for student behavior classification. The main objective of this study was to design, develop, and test a digital image–based exam cheating detection model capable of recognizing various types of behavior, such as looking around, glancing suspiciously, as well as non-cheating behaviors like focusing and boredom. The dataset used in this study consisted of 1,200 digital images categorized into six different behavior classes. The dataset was divided into three parts: 70% for training (840 images), 10% for validation (120 images), and 20% for testing (240 images). The experimental results showed that the SURF-CNN approach achieved better performance compared to the standard CNN. The SURF-CNN model achieved an accuracy of 91.80% on training data, 88.65% on validation, and 86.15% on testing, while CNN only achieved 88.20% on training, 85.25% on validation, and 83.15% on testing
Co-Authors Abd Rahman Imam Mulya Muchamad Sandy Pupu Ramadhan