cover
Contact Name
Hidayat
Contact Email
hidayat@email.unikom.ac.id
Phone
-
Journal Mail Official
tk@email.unikom.ac.id
Editorial Address
-
Location
Kota bandung,
Jawa barat
INDONESIA
KOMPUTIKA - Jurnal Sistem Komputer
ISSN : 22529039     EISSN : 26553198     DOI : -
Jurnal Ilmiah KOMPUTIKA adalah wadah informasi berupa hasil penelitian, studi kepustakaan, gagasan, aplikasi teori dan kajian analisis kritis di bidang kelimuan bidang Sistem Komputer.
Arjuna Subject : -
Articles 225 Documents
Hyperparameter Optimization of Random Forest for Multiclass Classification of Student Academic Performance Using Multidimensional Factors Sri Nurhayati; Diana Effendi; Bobi Kurniawan Soegoto; Adam Mukharil Bachtiar; Hanhan Maulana; Ednawati Rainarli
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.v15i1.18885

Abstract

Classification for academic performances among students in a multi-class scenario is a challenging task due to its dependencies on multiple factors and characteristics, particularly in the medium academic performance category. This scenario makes it a problem for some models with their conventional settings in terms of their ability to optimally distinguish categories of academic performances while being used in classification tasks, thus leading to the need for optimization techniques in enhancing their performances. This research paper will design an optimization strategy for improving the performances of the Random Forest algorithm in a multi-class academic performance classification among students. This will help in enhancing decision-making systems in education. The research method used is a machine learning approach with a Random Forest algorithm optimized through hyperparameter tuning using RandomizedSearchCV. This study utilizes secondary student data obtained from the Kaggle public repository, consisting of 6,607 data points with 20 determining factors covering academic, behavioral, social, environmental, and health aspects. The results showed that Random Forest hyperparameter optimization was able to improve model performance from a baseline accuracy of 79.56% to 81.08% on the validation data, and achieved an accuracy of 81.69% on the test data. In addition, there was an improvement in performance in the Medium category classification, as indicated by an increase in the F1-score value from 0.69 to 0.72. Therefore, the optimization of Random Forest proved to be good in enhancing the performance and stability of multiclass classification of student academic performance.
Smart Notification System with the Integration of Robotic Process Automation and Reinforcement Learning Andri Heryandi; Sufa Atin; Hani Irmayanti; Adam Mukharil Bachtiar; Hanhan Maulana; Bobi Kurniawan Soegoto; Ednawati Rainarli
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.v15i1.18951

Abstract

This study proposes the development of an intelligent academic notification system by integrating Robotic Process Automation (RPA) and Reinforcement Learning (RL) to improve the effectiveness of delivering information to students and parents. RPA is utilized to automate the process of sending notifications across various channels, such as email and WhatsApp, ensuring fast, consistent, and hands-free message distribution. RL is implemented to determine the optimal communication channel based on delivery history, message status (sent, failed, read), and the cost associated with each channel. Each student is represented as a state, while the selection of a communication channel becomes an action evaluated using Q-learning. The system learns from recipient behavior and updates the Q-table to enhance the accuracy of channel selection for future notifications. Additionally, the system applies an automatic escalation mechanism to parents as the deadline approaches. The result of this research is a smart notification system that can be implemented within academic information systems to enhance operational efficiency and student engagement.
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.
Comparative Analysis of Color-Based Thresholding and Thresholding-SVM Methods for Fire Image Classification Susmini Indriani Lestariningati; Mochamad Fajar Wicaksono; Myrna Dwi Rahmatya; Aldo Agusdian; Meita Maharani Iskandar
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.v15i1.19271

Abstract

Fire disasters remain a serious and recurring problem in Indonesia, particularly in densely populated urban areas and forest and land regions, causing significant material losses and posing serious threats to human safety. Early fire detection is therefore essential to minimize damage and casualties. Conventional fire detection systems suffer from limited coverage and delayed response, motivating the use of image-based fire detection as an alternative solution. Among image processing approaches, thresholding-based methods are widely used due to their simplicity and low computational cost. However, their performance may vary under different environmental conditions. To improve robustness, thresholding techniques are often combined with machine learning classifiers such as Support Vector Machines (SVM). This study presents a comparative analysis of fire image classification using standalone thresholding and thresholding combined with SVM. The dataset consisted of 999 fire and non-fire images collected from publicly available sources and was divided into training and testing sets using an 80:20 split. Experimental results show that the standalone thresholding method achieves an accuracy of 87.39%, outperforming the thresholding combined with SVM, which achieves an accuracy of 75.58%. Although the SVM classifier successfully identifies all fire images, it fails to distinguish non-fire images, resulting in a high false positive rate. These findings indicate that increased model complexity does not necessarily improve classification performance when feature representation is limited. These results provide practical insights into the effectiveness of lightweight fire image classification methods and highlight the importance of feature selection in machine learning-based fire detection systems.
Prediksi Penyakit Liver Dengan Model Pebandingan Logistic Regression, Support Vector Machine, Dan Random Forest Hidayat Hidayat; Mochamad Fajar Wicaksono; Yudis Jalu Wicaksono; Sandy Maulana Wibawa; Edvan Nuriana Putra Septia
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.v15i1.19799

Abstract

Penyakit liver sering berkembang tanpa gejala awal yang jelas sehingga menyulitkan proses deteksi dini dan meningkatkan risiko komplikasi serius. Keterbatasan analisis manual terhadap data laboratorium pasien mendorong perlunya pendekatan komputasi yang mampu memberikan klasifikasi penyakit secara lebih akurat dan konsisten. Pendekatan pembelajaran mesin diterapkan untuk mengklasifikasikan potensi penyakit liver menggunakan data klinis terstruktur. Kontribusi utama yang dihasilkan berupa evaluasi komprehensif tiga algoritma klasifikasi dengan pengujian internal serta validasi eksternal menggunakan data yang tidak terlibat dalam pelatihan. Dataset yang digunakan terdiri dari 579 data pasien yang telah dibersihkan dan dibagi menjadi data pengembangan serta data uji akhir. Pemodelan dilakukan menggunakan algoritma regresi logistik, hutan acak, dan mesin vektor pendukung dengan variasi pembagian data latih sebesar 60%, 70%, dan 80% melalui pengambilan sampel acak berulang. Hasil pengujian menunjukkan bahwa metode hutan acak menghasilkan performa terbaik dengan nilai F1 sebesar 0,696 dan koefisien korelasi Matthews sebesar 0,203 pada pembagian data 80%. Regresi logistik menunjukkan akurasi stabil sebesar 0,728 namun memiliki kemampuan generalisasi yang lebih rendah. Mesin vektor pendukung menghasilkan akurasi cukup tinggi tetapi menunjukkan ketidakseimbangan klasifikasi. Pengujian data baru menunjukkan bahwa metode hutan acak mencapai kecocokan prediksi sebesar 98%, tertinggi di antara seluruh metode. Hasil tersebut menunjukkan bahwa metode hutan acak lebih efektif dalam klasifikasi penyakit liver karena memberikan keseimbangan antara akurasi, stabilitas, dan kemampuan generalisasi model.