Claim Missing Document
Check
Articles

Found 25 Documents
Search

AI-Powered Face Mask Detection Utilizing MobileNetV2 for Health Monitoring Misinem Misinem; Eka Puji Agustini; Maria Ulfa
International Journal of Advances in Artificial Intelligence and Machine Learning Vol. 1 No. 1 (2024): International Journal of Advances in Artificial Intelligence and Machine Learni
Publisher : CV Media Inti Teknologi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.58723/ijaaiml.v1i1.286

Abstract

The COVID-19 pandemic has highlighted the critical need for face masks to prevent virus transmission. Ensuring consistent mask usage in crowded public spaces remains a challenge, especially with manual monitoring methods that are inefficient and prone to error. To address this, this research introduces a real-time face mask detection system leveraging MobileNet-V2, a lightweight and efficient deep learning model known for its high performance in image classification tasks. The system utilizes a dataset from Kaggle comprising 11,792 images, divided into training (10,000), validation (800), and testing (992) sets. MobileNet-V2 was fine-tuned for this task, using its inverted residual layers to extract features and enhance performance effectively. Data augmentation techniques were applied to improve the model’s ability to generalize across diverse scenarios. The MobileNet-V2 model achieved an impressive 98.69% accuracy on the testing dataset, demonstrating exceptional reliability in identifying individuals wearing masks versus those without. Standard evaluation metrics, including precision, recall, and a confusion matrix, confirmed its robustness. This system’s ability to operate in real-time makes it ideal for public health surveillance in environments such as airports, shopping malls, and public transport. The proposed face mask detection system is both accurate and scalable, offering an efficient solution for enforcing mask-wearing protocols in public spaces. The system’s integration of advanced deep learning techniques ensures its reliability in real-time monitoring, contributing to better public health management. Future work will focus on further optimizing the model and expanding its application to other health-related monitoring tasks, enhancing its value for public health surveillance.
Real-Time Outlier Detection in Fast-Moving Data Streams Eka Puji Agustini; Mohd Zaki Zakaria
International Journal of Advances in Artificial Intelligence and Machine Learning Vol. 1 No. 1 (2024): International Journal of Advances in Artificial Intelligence and Machine Learni
Publisher : CV Media Inti Teknologi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.58723/ijaaiml.v1i1.287

Abstract

Anomaly detection is a critical task in various fields such as finance, healthcare, network monitoring, and sensor data analysis, where identifying unusual patterns or outliers in data streams is essential for timely decision-making. Two commonly used techniques for anomaly detection are the Moving Average (MA) and Exponential Moving Average (EMA) methods. Despite their widespread use, selecting the appropriate method depends on the nature of the data and the requirements of the system. This paper presents a comparative analysis of MA and EMA for anomaly detection, focusing on critical factors such as speed of detection, stability, precision and recall, false positive rate, and computational efficiency. This research addresses the problem of determining which method, MA or EMA, is better suited for specific types of data, particularly in streaming environments with varying trends and anomalies. The results of our comparison indicate that EMA performs better in dynamic environments where rapid identification of anomalies is critical, such as financial markets or network traffic analysis. It quickly detects sudden deviations but may flag minor fluctuations as false positives due to its sensitivity. MA, on the other hand, is more stable and computationally efficient, with a lower false positive rate, making it more suitable for applications where long-term trend monitoring is required, and stability is prioritized over speed. This research highlights the strengths and weaknesses of both methods, demonstrating that the choice between MA and EMA should be based on the specific needs of the anomaly detection system. For real-time, high-speed environments, EMA offers a more responsive solution, while MA provides better stability and efficiency in long-term monitoring. A hybrid approach combining both methods could offer a more robust solution, adapting to different types of data and detection requirements.
ANALISIS KUALITAS WEBSITE PEMERINTAH DAERAH MENGGUNAKAN METODE WEB QUAL 4.0 DAN IMPORTANCE PERFORMANCE ANALYSIS Febriyanti Panjaitan; Rahjan Saputra; Susan Dian Purnamasari; Ch. Desi Kusmindari; Eka Puji Agustin Puji Agustini
JURNAL PERANGKAT LUNAK Vol 5 No 2 (2023): Jurnal Perangkat Lunak
Publisher : Indragiri Islamic University

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.32520/jupel.v5i2.2539

Abstract

Dalam era digital saat ini, teknologi informasi menjadi semakin penting bagi pemerintah untuk memenuhi kebutuhan masyarakat. Salah satu bentuk pelayanan publik melalui teknologi informasi adalah melalui website pemerintah daerah. Namun, masih ada beberapa website pemerintah daerah yang belum memenuhi standar kualitas yang baik dalam hal informasi yang disajikan dan kemudahan akses pengguna. Oleh karena itu, penelitian ini bertujuan untuk menganalisis kualitas website pemerintah daerah di kabupaten Ogan Komering Ilir (OKI) menggunakan metode Web Qual 4.0 dan Importance Performance Analysis (IPA). Metode ini telah digunakan dalam beberapa penelitian sebelumnya. Dalam penelitian ini, hasil analisis menunjukkan bahwa kepuasan pengguna dalam menggunakan website pemerintah cukup puas untuk 81% dari semua indikator, sedangkan harapan pengguna dengan hasil analisis kesesuaian sebesar 80% dan rata-rata hasil kesenjangan -0.81. Rekomendasi perbaikan dapat diberikan terutama pada atribut IQ3 (Menyediakan informasi yang up to date) dan atribut SIQ3 (Rasa aman dalam menyampaikan data pribadi) pada kuadran I untuk meningkatkan kualitas pelayanan publik melalui website
Self-Supervised Learning Strategies for Reducing Label Dependency in Large-Scale Data Science Applications Kit Ling Chan; Mohd Zaki Zakaria; Eka Puji Agustini
International Journal of Advances in Artificial Intelligence and Machine Learning Vol. 3 No. 2 (2026): International Journal of Advances in Artificial Intelligence and Machine Learni
Publisher : CV Media Inti Teknologi

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.58723/ijaaiml.v3i2.802

Abstract

Background: The rapid expansion of large-scale data ecosystems has intensified the challenge of acquiring high-quality labeled datasets for supervised learning models. Although raw data availability continues to grow exponentially, manual annotation remains costly, time-consuming, and susceptible to inconsistency, creating a structural bottleneck in scalable machine learning deployment. Existing self-supervised learning (SSL) approaches have demonstrated promising results in domain-specific applications, particularly in computer vision. However, many lack a unified, modality-agnostic framework and systematic evaluation under realistic constraints such as limited-label availability and noisy annotations.Aims: This study aims to develop a unified self-supervised learning framework that reduces label dependency across heterogeneous data modalities. The scope of the research includes image and tabular datasets, focusing on evaluating model robustness under low-label conditions and label noise scenarios.Methods: The proposed framework integrates contrastive representation learning with reconstruction-based objectives in a two-stage training strategy. First, large-scale unlabeled pre-training is conducted to learn robust feature representations. This is followed by supervised fine-tuning under controlled label fractions (1%–20%) and synthetic label corruption levels of up to 30%. Performance is evaluated against fully supervised baselines using accuracy and data efficiency metrics.Result: Experimental findings demonstrate that the proposed SSL framework achieves up to 15% higher accuracy than fully supervised baselines under 1% labeled data conditions. Moreover, it maintains approximately 85–90% of clean-label performance even with 30% label corruption. The data efficiency ratio improves by up to 2.3× in low-label regimes, indicating substantial gains in learning effectiveness under constrained supervision.Conclusion: The results confirm that self-supervised pre-training significantly enhances representation quality, robustness, and scalability while reducing annotation requirements. This research establishes SSL as a foundational paradigm for developing data-efficient and resilient AI systems in large-scale analytics environments.
Predicting Student Loyalty in Higher Education Using Machine Learning: A Random Forest Approach Qoriani Widayati; Kusworo Adi; R Rizal Isnanto; Eka Puji Agustini; Dewa Rizki Rahmat Julianto; Fawwaz Bimo Prakasa
Journal of Information System and Informatics Vol 7 No 1 (2025): March
Publisher : Asosiasi Doktor Sistem Informasi Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.51519/journalisi.v7i1.977

Abstract

Student loyalty is a crucial factor supporting the sustainability of higher education institutions. The aim of this study is to predict student loyalty using a machine learning approach, specifically the random forest algorithm. The data for this research were collected through a questionnaire that included variables such as service quality, emotional attachment, brand satisfaction, brand trust, and socio-economic conditions, distributed to 107 students in Palembang. The resulting dataset was processed through preprocessing, model training, and performance evaluation, employing metrics such as accuracy, precision, recall, and F1-score. The analysis using the random forest algorithm achieved an accuracy of 90.9%. These findings are expected to provide valuable insights for higher education institutions in developing more effective strategies to enhance student loyalty.