cover
Contact Name
Ardi Susanto
Contact Email
ardisusanto@poltektegal.ac.id
Phone
-
Journal Mail Official
informatika.ejournal@poltektegal.ac.id
Editorial Address
Gedung B, Politeknik Harapan Bersama, Jl Mataram No 9 Pesurungan Lor Kota Tegal
Location
Kota tegal,
Jawa tengah
INDONESIA
Jurnal Informatika: Jurnal Pengembangan IT
ISSN : 24775126     EISSN : 25489356     DOI : https://doi.org/10.30591
Core Subject : Science,
The scope encompasses the Informatics Engineering, Computer Engineering and information Systems., but not limited to, the following scope: 1. Information Systems Information management e-Government E-business and e-Commerce Spatial Information Systems Geographical Information Systems IT Governance and Audits IT Service Management IT Project Management Information System Development Research Methods of Information Systems Software Quality Assurance 2. Computer Engineering Intelligent Systems Network Protocol and Management Robotic Computer Security Information Security and Privacy Information Forensics Network Security Protection Systems 3. Informatics Engineering Software Engineering Soft Computing Data Mining Information Retrieval Multimedia Technology Mobile Computing Artificial Intelligence Games Programming Computer Vision Image Processing, Embedded System Augmented/ Virtual Reality Image Processing Speech Recognition
Articles 487 Documents
Analisis Efektivitas Fine-Tuning dan Prompt Engineering Berbasis Llama 3.1 pada Deteksi Depresi di Media Sosial Muhammad Ikhsan Asagaf; Junta Zeniarja
Jurnal Informatika: Jurnal Pengembangan IT Vol 11, No 2 (2026)
Publisher : Politeknik Harapan Bersama

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30591/jpit.v11i2.10170

Abstract

Detecting depression has become an important concern in addressing mental health issues. According to WHO, more than 300 million people suffer from depression. Large Language Models offer great potential to address this issue, however the full fine-tuning process is often hampered by heavy computational requirements, and LLMs that are not specifically configured for a particular context can result in biased and inaccurate outcomes. This study aims to analyze the effectiveness of Prompt Engineering and Fine-Tuning using QLoRA in improving the accuracy of depression detection. Utilizing the Llama-3.1-8B-Instruct model on social media datasets, this research compares model performance in two scenarios consist of the application of direct prompting strategies on the base model and the application of QLoRA fine-tuning. Evaluation results demonstrate that the Chain-of-Thought strategy improved baseline accuracy from 81.4% to 84.4%, but still exhibited significant bias towards the 'Severe' class. In contrast, the QLoRA Fine-Tuning approach proved superior, achieving 92.4% accuracy with balanced F1-Scores across classes, effectively eliminating detection bias in the 'Minimum' class. These findings confirm that while prompting techniques can enhance baseline performance, QLoRA provides a more accurate, stable, and objective solution for depression detection tasks.
Deteksi Edema Paru Pada Citra Chest X-ray Menggunakan YOLOv5n Dengan Optimasi Hyperparameter Berbasis Grey Wolf Optimizer Nasrudin Affandi Prasetyo; Cinantya Paramita; Amiq Fahmi
Jurnal Informatika: Jurnal Pengembangan IT Vol 11, No 2 (2026)
Publisher : Politeknik Harapan Bersama

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30591/jpit.v11i2.10375

Abstract

Pulmonary edema is a lung disorder characterized by fluid accumulation in the alveolar and interstitial spaces, which disrupts gas exchange and reduces oxygen levels in the blood. Chest X-ray (CXR) imaging is commonly used for pulmonary edema assessment because it is fast and widely available; however, its interpretation still depends heavily on radiologist expertise and may lead to diagnostic variability, particularly in healthcare facilities with limited radiology resources. This study aims to develop an automated pulmonary edema detection system based on deep learning to support more consistent analysis of CXR images. The novelty of this study lies in the integration of Grey Wolf Optimizer (GWO) for hyperparameter optimization of the YOLOv5n model specifically for pulmonary edema detection in Chest X-ray images. The proposed method employs YOLOv5n as a lightweight object detection architecture because of its computational efficiency and suitability for resource-constrained environments. To improve detection performance and training stability, the hyperparameters of YOLOv5n are optimized using GWO. The model is trained and evaluated using annotated CXR images, and its performance is measured using precision, recall, mAP@0.5, and mAP@0.5–0.95. Experimental results show that the YOLOv5n + GWO model achieved a precision of 0.906, recall of 0.936, mAP@0.5 of 0.963, and mAP@0.5–0.95 of 0.737, indicating improved detection performance compared with the baseline YOLOv5n configuration. The proposed framework demonstrates potential as a decision-support tool for assisting medical personnel in early pulmonary edema screening through efficient and consistent analysis of CXR images.
Penerapan Fuzzy Sugeno untuk Deteksi Overload Host pada Dynamic VM Consolidation M Naufal Adrian Pratama Putra; Chaerul Umam
Jurnal Informatika: Jurnal Pengembangan IT Vol 11, No 2 (2026)
Publisher : Politeknik Harapan Bersama

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30591/jpit.v11i2.10130

Abstract

High energy consumption in cloud data centers has a direct impact on operational costs and environmental efficiency, so a more adaptive and energy-efficient resource management strategy is needed through Dynamic Virtual Machine (VM) consolidation. This research proposes a novelty in the form of applying the Zero Order Fuzzy Sugeno method as a more responsive and lightweight host overload detection mechanism. The proposed method was tested through simulations using CloudSim version 7 with PlanetLab workloads. Performance evaluation was conducted through simulations using the CloudSim 7G framework with a standard PlanetLab benchmark workload. Tests were executed in 12 cross-combination scenarios to compare the performance of Fuzzy Sugeno with the system's built-in benchmarking methods, such as Inter Quartile Range (IQR) and Median Absolute Deviation (MAD). The test results show that the Fuzzy Sugeno method is able to provide significant energy consumption savings of up to 27.5%, with the lowest energy consumption achievement of 144.07 kWh, much more efficient than the IQR-MU comparison method which reached 198.94 kWh. Although there is a trade-off in the form of an increase in the frequency of VM migration which results in a violation of the Service Level Agreement (SLA) of 0.70%, the Fuzzy Sugeno method has proven to be very effective and is recommended for green data center scenarios that prioritize maximum power efficiency.
Analisis Korelasi Curah Hujan dan Lahan Terbangun Terhadap Luasan Genangan Banjir di Kabupaten Demak Mikael Arvito Kurnia Adi; Agusta Praba Ristadi Pinem; Nurtriana Hidayati
Jurnal Informatika: Jurnal Pengembangan IT Vol 11, No 2 (2026)
Publisher : Politeknik Harapan Bersama

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30591/jpit.v11i2.10392

Abstract

Demak Regency is a low-lying coastal area that faces two types of pressures due to natural phenomena and massive human activities. The purpose of this study is to examine changes in built-up land and measure the level of spatial relationship between rainfall and built-up land area with flood inundation area in Demak Regency during the period 2020 to 2024. The method used in this study is spatial quantitative based on cloud computing using the Google Earth Engine platform that utilizes CHIRPS data for rainfall, Dynamic World for built-up land, and Sentinel-1 for flood inundation area. The findings of this study indicate that the annual rainfall trend has a low level of correlation with a coefficient of determination ranging from 1% to 7% while changes in built-up land show a high level of correlation with a Pearson correlation value between -0.52 to -0.58 which indicate that the reduction of non-built-up land due to land conversion is directly propotional  to the expansion of inundation which contributes to the variation of flood inundation by around 34%. The increase in flood inundation area reaching ±6,970 Ha in 2024 amid normal rainfall confirms that the decline in environmental capacity occurs due to excessive land conversion. This study concludes that managing the risk of increased flooding in Demak Regency requires strict integration of spatial planning policies, not just relying on a meteorology-based early warning system.
Perbandingan Performa CNN, Bi-LSTM, dan LSTM Untuk Analisis Sentimen Komentar Berbahasa Melayu Bengkulu Julia Purnama Sari; Funny Farady Coastera; Niska Ramadani; Esti Asmareta Ayu
Jurnal Informatika: Jurnal Pengembangan IT Vol 11, No 2 (2026)
Publisher : Politeknik Harapan Bersama

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30591/jpit.v11i2.10376

Abstract

Bengkulu Malay is an under-resourced regional language used in daily digital communication, including public comments on Instagram and TikTok. Its informal vocabulary, spelling variation, and limited language resources make sentiment classification more challenging than for standard Indonesian. This study compares Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), and Bidirectional Long Short-Term Memory (BiLSTM) for sentiment analysis of Bengkulu Malay comments. The dataset was collected from three local issues, namely Q1, Q2, and Q3, with 22,888 raw comments and 3,961 labeled comments after selection. The research pipeline includes data collection, labeling, text preprocessing adapted to Bengkulu Malay, Word2Vec Skip-gram embedding, data splitting into training, validation, and testing sets, model training, and evaluation using accuracy, precision, recall, F1-score, and confusion matrix. The final labeled dataset was split into 80% training and 20% testing data, and 20% of the training data was used for validation. The results show that BiLSTM achieved the best accuracy on all datasets, with 91% on Q1, 85% on Q2, and 70% on Q3. CNN achieved 69%, 60%, and 64%, while LSTM achieved 49%, 55%, and 64%. These results indicate that BiLSTM is more effective in capturing bidirectional context in Bengkulu Malay comments, although performance decreases when class imbalance becomes stronger.
Anomaly-Based Network Intrusion Detection Using Isolation Forest on the Imbalanced UNSW-NB15 Dataset Syifaurachman Syifaurachman; Samso Supriyatna; Muhamad Ihsan Ashari
Jurnal Informatika: Jurnal Pengembangan IT Vol 11, No 2 (2026)
Publisher : Politeknik Harapan Bersama

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30591/jpit.v11i2.10378

Abstract

 Modern cyber threats require adaptive intrusion detection systems (IDS) capable of identifying complex anomalies within large networks. . Traditional signature-based methods frequently struggle to detect novel zero-day attacks, making unsupervised machine learning, specifically Isolation Forest (IF), a vital research direction. IF detects anomalies by isolating unusual observations through random partitioning mechanisms.. This study evaluates IF for intrusion detection using the highly imbalanced UNSW-NB15 dataset, comprising 2,540,044 records (87% normal traffic, 13% attack traffic). A quantitative approach was applied, evaluating performance across Accuracy, Precision, Recall, F1-score, ROC-AUC, and PR-AUC. Results show an accuracy of 79.3% and a ROC-AUC of 0.6302. However, a low PR-AUC (0.1710) and F1-score (0.18) reveal the model's critical failure in detecting the minority attack class due to severe feature overlap. These findings prove that relying strictly on Accuracy or ROC-AUC is misleading in imbalanced scenarios. Therefore, IF should serve only as a lightweight baseline screening layer; hybrid architectures are necessary to improve detection sensitivity in real-world IDS.
Adaptasi Profil Proyeksi Horisontal dengan Stripping Vertikal untuk Segmentasi Baris Teks Miring Arab Jawi Ade Jamal; Iin Suryaningsih; Arif Supriyanto
Jurnal Informatika: Jurnal Pengembangan IT Vol 11, No 2 (2026)
Publisher : Politeknik Harapan Bersama

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30591/jpit.v11i2.10085

Abstract

Abstract – Line segmentation in handwritten Arab Jawi scripts poses a significant challenge for automatic character recognition systems due to high variations in writing slant. The primary problem arises when using the standard Horizontal Projection Profile (HPP) method, where text skewness causes an overlap between peaks and valleys in the projection graph, leading to inaccurate or failed line separation. This research aims to develop a line segmentation solution for skewed Arab Jawi text by adapting the HPP method through a vertical stripping technique. The methodology involves dividing the document image into several narrow vertical columns or strips, followed by independent horizontal projection calculations for each strip. Local line separation points from each strip are then sequentially connected to form a dynamic separation path that follows the original inclination angle of the text. Research findings demonstrate that this approach successfully separates skewed text lines perfectly without cutting through characters, while also effectively managing variations in manuscript colour degradation and inconsistent line sizes. In conclusion, the modification of HPP with vertical stripping proves to be effective and computationally efficient as a pre-processing stage for ancient Jawi manuscripts. This method offers a balance between the simplicity of classical algorithms and the robustness required to handle handwriting complexity, making it highly potential for integration into the development of broader Optical Character Recognition systems to support the preservation of historical Southeast Asian manuscripts
Perbandingan Algoritma C4.5 dan Random Forest dalam Klasifikasi Kekeringan Tembakau Berbasis Electronic Nose Rahmat Tegar Patriot Hari Lambang; Dimas Saputra; Muh. Mashdarul Hilmi Aufa; Ifnu Wisma Dwi Prastya
Jurnal Informatika: Jurnal Pengembangan IT Vol 11, No 2 (2026)
Publisher : Politeknik Harapan Bersama

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30591/jpit.v11i2.10057

Abstract

Tobacco quality is influenced by changes in volatile compounds that occur during the drying process and can be represented through the sensor response patterns on the electronic nose (E-Nose) device. Various machine learning algorithms have been used to classify tobacco based on E-Nose data, but the performance of each algorithm can vary depending on the characteristics of the data used. Therefore, this study aims to analyze and compare the performance of the C4.5 and Random Forest algorithms in classifying tobacco quality based on volatile compound data obtained using the E-Nose. A total of 375 tobacco sample data representing four drying conditions were used in this study. The preprocessing stage was carried out using the Interquartile Range (IQR) method to remove outliers and Moving Average to reduce noise in the MQ-4, MQ-7, and MQ-135 sensor signals. Furthermore, both algorithms were evaluated using stratified 10-fold cross-validation to obtain stable performance estimates. The results indicated that Random Forest performed better than C4.5 in all evaluation metrics. Random Forest achieved an accuracy of 94.44%, a Cohen's Kappa value of 0.9259, an MCC of 0.9262, a balanced accuracy of 0.945, and a cross-entropy log loss of 0.4291. These results indicate that Random Forest is more effective in classifying tobacco quality based on volatile compound data obtained using E-Nose.
Comparative Evaluation of VGG16, MobileNetV2, and ResNet50 for Pediatric Pneumonia Classification Using Grad-CAM Rendra Gunawan; Cinantya Paramita; Suryanti Suryanti
Jurnal Informatika: Jurnal Pengembangan IT Vol 11, No 2 (2026)
Publisher : Politeknik Harapan Bersama

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30591/jpit.v11i2.10373

Abstract

 Pneumonia remains one of the leading causes of mortality among children worldwide, particularly in developing countries, where early and accurate diagnosis is crucial. This study aims to evaluate and compare the performance of three deep learning architectures, namely VGG16, MobileNetV2, and ResNet50, for pediatric pneumonia classification using chest X-ray images. The dataset consists of pediatric chest radiographs (ages 1-5 years) obtained from Guangzhou Women and Children’s Medical Center, which were preprocessed through normalization and data augmentation techniques to improve model generalization. The classification task involves three categories: normal, bacterial pneumonia, and viral pneumonia. Model performance was evaluated using accuracy, precision, recall, specificity, F1-score, G-Mean, and AUC. The experimental results show that all models achieve competitive performance, with accuracy ranging from 79% to 82%, where VGG16 outperforms the other architectures. Furthermore, Grad-CAM is applied to enhance interpretability by visualizing important regions in X-ray images that influence model decisions. The results demonstrate that Grad-CAM provides meaningful visual explanations, supporting the reliability of deep learning models in assisting clinical diagnosis.
Segmentasi Kepuasan Mahasiswa Terhadap Dosen Menggunakan K-Means Clustering dan Identifikasi Faktor Dominan Dengan Random Forest Restu Rakhmawati; Suamanda Ika Novichasari; Imam Adi Nata; Fadhila Syahida Wibowo; Zharifa Nur Majidah; Meily Adenia
Jurnal Informatika: Jurnal Pengembangan IT Vol 11, No 2 (2026)
Publisher : Politeknik Harapan Bersama

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.30591/jpit.v11i2.10395

Abstract

This study aims to analyze student satisfaction patterns regarding lecturer teaching performance by integrating K-Means Clustering and Random Forest algorithms. The research data includes 5,260 observations analyzed based on service quality dimensions and academic attributes. The results using the Elbow Method and Silhouette Score established three optimal clusters representing Very Satisfied (score 4.84), Moderately Satisfied (score 4.04), and Dissatisfied (score 3.04) segments. Furthermore, the Random Forest algorithm demonstrated an accuracy of 47% on 1,576 test data and successfully identified that Semester Credit Load (SKS) is the most dominant determinant influencing satisfaction, with an importance value of 47.46%. A unique finding shows that students with the highest academic load (average 20.81 SKS) are actually in the Very Satisfied segment. This study concludes that more intensive student academic engagement correlates positively with appreciation for lecturer teaching quality. These results provide strategic guidance for university management to improve services based on student academic profiles.