Article Per Year (5 Year)
p-Index From 2021 - 2026
0.408
P-Index
This Author published in this journals
All Journal Jurnal Sarjana Teknik Informatika JUITA : Jurnal Informatika
Nurhadi Nurhadi
Institut Teknologi Dan Bisnis Riau Pesisir
Computer Science & IT

Published : 2 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 2 Documents
Search

 1 
Stacking-Based Hybrid Ensemble Learning for Security Personnel Attendance Prediction and Performance Classification Nurhadi Nurhadi; Khairul Azmi Azmi; Mustazzihim Suhaidi Suhaidi; Sandi Fadilah Fadilah
Jurnal Sarjana Teknik Informatika Vol. 14 No. 2 (2026): Juni
Publisher : Program Studi Informatika, Universitas Ahmad Dahlan

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.12928/jstie.v14i2.32090

Abstract

Monitoring the attendance and performance of campus security personnel is essential to ensure operational effectiveness and service reliability. However, conventional monitoring approaches are often manual, fragmented, and prone to inaccuracies, limiting timely decision-making. This study proposes a stacking-based hybrid ensemble machine learning model for predicting security personnel attendance and classifying performance within the SIMSATPAM campus monitoring system at Institut Teknologi dan Bisnis Riau Pesisir. The proposed method integrates Random Forest, Support Vector Machine (SVM), and XGBoost as base learners, while Logistic Regression is employed as the meta-learner to improve predictive capability and classification robustness. To prevent overfitting and data leakage, the stacking architecture utilizes K-Fold Cross-Validation and Out-of-Fold (OOF) prediction mechanisms during meta-learner training. Experimental results demonstrate that the proposed stacking hybrid ensemble model outperforms individual classifiers across all evaluation metrics. The model achieved an accuracy of 94.27%, precision of 93.81%, recall of 93.45%, and F1-score of 93.62%, improving accuracy by 4.13% compared to Random Forest, 5.51% compared to SVM, and 2.24% compared to XGBoost. Furthermore, the proposed model produced stable multi-class classification performance for attendance prediction and personnel performance evaluation. Analysis indicates that attendance duration, shift schedule, and lateness frequency are the most influential variables affecting prediction outcomes. These findings confirm that the proposed stacking-based hybrid ensemble approach provides an effective intelligent decision-support system for automated campus security monitoring and personnel management in smart campus environments.
Edge AI-Based Multimodal Biometric Smart Reader Using YOLOv8 for Integrated Academic Attendance Systems Nurhadi Nurhadi; Emil Naf'an; Desyanti Desyanti; Mustazzihim Suhaidi
JUITA: Jurnal Informatika JUITA Vol. 14 Issue 2, July 2026
Publisher : Department of Informatics Engineering, Universitas Muhammadiyah Purwokerto

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30595/juita.v14i2.30527

Abstract

Attendance systems in vocational education institutions face challenges related to accuracy, security, and susceptibility to manipulation due to the use of single-modality authentication methods. RFID-based systems are vulnerable to card sharing, fingerprint systems suffer from latency during peak usage, and face recognition systems are sensitive to illumination and pose variations. This study proposes an Edge AI-based multimodal biometric smart reader integrating RFID, fingerprint, and YOLOv8-based face recognition for an academic attendance system at SMK Negeri 1 Dumai. The system is implemented on NVIDIA Jetson Nano as an edge computing device and integrated with an academic information system through an IoT-based architecture for real-time attendance monitoring. A decision-level fusion approach using majority voting is applied, where authentication is accepted if at least two of three modalities match. The system is evaluated using accuracy, False Acceptance Rate (FAR), False Rejection Rate (FRR), and response time. Experimental results show that the proposed multimodal system achieves an accuracy of 98.72%, outperforming RFID (89.34%), fingerprint (92.15%), and YOLOv8 face recognition (95.63%). The system also reduces FAR to 0.82% and FRR to 0.91%, with an average response time of 1.47 seconds, making it suitable for real-time deployment. Overall, the proposed Edge AI-based multimodal biometric system demonstrates high accuracy, improved security, and efficient real-time performance, providing a scalable solution for intelligent attendance systems in vocational education environments.
Co-Authors Desyanti Desyanti Emil Naf'an Khairul Azmi Azmi Mustazzihim Suhaidi Mustazzihim Suhaidi Suhaidi Sandi Fadilah Fadilah