Article Per Year (5 Year)
p-Index From 2021 - 2026
0.408
P-Index
This Author published in this journals
All Journal KOMPUTIKA - Jurnal Sistem Komputer Bulletin of Network Engineer and Informatics (BUFNETS)
Muhammad Rifky Afandi
Universitas Pendidikan Indonesia
Computer Science & IT Control & Systems Engineering Electrical & Electronics Engineering

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

Found 2 Documents
Search

 1 
Visual Trend Analysis of E-Commerce Thumbnails Using Parallel Computing for Image Big Data Muhamad Tio Ariyanto; Haris Maulana; Muhammad Rifky Afandi; Eddy Prasetyo Nugroho
Komputika : Jurnal Sistem Komputer Vol. 15 No. 1 (2026): Komputika: Jurnal Sistem Komputer
Publisher : Computer Engineering Departement, Universitas Komputer Indonesia

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.34010/komputika.v15i1.19194

Abstract

The rapid growth of e-commerce platforms has led to the massive accumulation of product thumbnail images, making manual visual analysis inefficient and conventional sequential processing methods insufficient to handle such data volumes in a timely manner. Given the crucial role of thumbnails in influencing consumer purchasing decisions, computational strategies are required to accelerate the analysis process without compromising classification accuracy. This study applies a parallel computing approach combined with deep learning to improve the efficiency of visual trend analysis using two primary datasets: 2,608 images for model training and validation, and 40,254 images for large-scale inference. The proposed framework integrates parallel image preprocessing on multi-core CPUs, the development of a Convolutional Neural Network based on MobileNetV2 using a transfer learning approach, and batch-based parallel inference on GPUs. The developed model demonstrates stable and convergent performance, achieving a training accuracy of 0.85 and a validation accuracy of 0.83. Efficiency testing during the preprocessing stage shows that the parallel approach is more effective under large data workloads, providing a speed improvement of up to 1.58×. During the inference stage, predictions for 500 images can be completed in 1.84 seconds compared to 41.76 seconds using the sequential method, resulting in a significant computational speedup of 22.8×. Big data analysis reveals a polarization of visual strategies, where technology product categories are dominated by infographic-style thumbnails, fashion categories rely heavily on human model representations, and household product categories emphasize clean product visuals supported by promotional elements. This study concludes that the application of parallel computing significantly enhances the efficiency and scalability of visual big data analysis in e-commerce and supports more operational and strategic mapping of visual trends.
Transformer-Based IOC Classification: A Comparative Analysis of BERT Architectures and Traditional Machine Learning Baselines Muhammad Rifky Afandi; Rizky Rachman Judhie Putra; Eddy Prasetyo Nugroho
Bulletin of Network Engineer and Informatics Vol. 4 No. 2 (2026): BUFNETS (Bulletin of Network Engineer and Informatics) October 2026
Publisher : PT. GWEX NET PUBLISHER

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.59688/f7s1r692

Abstract

Open-source threat intelligence platforms provide vast amounts of Indicator of Compromise (IoC) data, but they are frequently saturated with benign operational noise and false alarms, causing severe alert fatigue in Security Operations Centres (SOCs). To mitigate this operational bottleneck, this study presents a comprehensive comparative analysis of automated IoC classification pipelines, evaluating traditional machine learning baselines against Transformer-based architectures. A dataset of 10,000 weakly labelled IoC records was utilised to compare Logistic Regression, Random Forest, and XGBoost (utilising TF-IDF n-grams) with DistilBERT and BERT-Base under two distinct scenarios: Zero-Shot Classification and Supervised Fine-Tuning. Empirical results revealed that while traditional tree-based models like Random Forest achieved a high raw accuracy of 93.78%, general-purpose Transformers in a Zero-Shot setting suffered from catastrophic predictive biases and failed entirely to filter false alarms. However, the application of Supervised Fine-Tuning completely rectified these limitations. The fine-tuned BERT-Base achieved a superior Weighted F1-Score of 92.74%, demonstrating exceptional proficiency in handling the severe class imbalance. Furthermore, an ablation study confirmed that the fine-tuned embeddings successfully learned the underlying syntactic structures of malicious indicators without relying on metadata shortcut learning. Finally, computational scalability tests proved that the lightweight DistilBERT architecture offers the most optimal balance between predictive accuracy and processing latency, rendering it highly viable for integration into batch-processing workflows in real-world SOC environments.
Co-Authors Eddy Prasetyo Nugroho Haris Maulana Muhamad Tio Ariyanto Rizky Rachman Judhie Putra