Scientific Publications Directory

Full Author List

Maharani.S, Putri; Satra, Ramdan, Universitas Muslim Indonesia; Azis, Huzain, Universiti Kuala Lumpur

Abstract

Distributed Denial of Service (DDoS) attacks pose a serious risk to network reliability, particularly within Software-Defined Networking (SDN) architectures that rely on centralized control. This research analyzes the effectiveness of Gaussian Naïve Bayes (GNB) and Support Vector Machine (SVM) with a Radial Basis Function (RBF) kernel for identifying DDoS attacks using an SDN traffic dataset consisting of 104,345 flow records. Three different feature-handling strategies are explored: using the complete feature set without reduction, applying feature selection through SelectKBest, and performing dimensionality reduction with Principal Component Analysis (PCA). Model validation is carried out using Stratified K-Fold Cross-Validation with K values of 2, 5, and 10. Predictions obtained from each fold are merged into a single aggregated confusion matrix to compute evaluation metrics, including accuracy, precision, recall, and F1-score. Experimental results demonstrate that RBF-SVM without feature reduction delivers the highest detection performance, reaching an accuracy of up to 96.9%, while GNB provides lower accuracy but operates with greater computational efficiency. These findings indicate that an evaluation framework based on aggregated confusion matrices can provide more dependable performance estimates for DDoS detection systems deployed in SDN environments.

Related SDGs Targets

Full Author List

Herman; Putra Muhammad Dani Arya, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia; Nasir, Haidawati, MIIT University Kuala Lumpur, Malaysia; Darwis, Herdianti; Mansyur, St. Hajrah, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia; Noor, Megat Nurolazmi MM, MIIT University Kuala Lumpur, Malaysia

Abstract

This study focuses on being able to classify traditional Toraja carvings using Convolutional Neural Networks (CNN), focusing on three CNN architectures, specifically VGG-16, VGG-19, and a model that is concatenated from both. The aim is to determine the most effective architecture and ratio of training data and validation data sharing to achieve the highest classification accuracy. The image dataset consisting of seven different Toraja carving motifs or classes underwent data pre-processing, namely data augmentation, to improve model generalization and reduce overfitting. Experiments were conducted using four scenarios of training data and validation data separation. The final outcome of this research is that VGG-16 reached the best validation performance of 97.36% with a data 90%: 10% separation. It manifests its superior ability to Capture the information of complicated Toraja carving motifs. VGG-19 and the combined model also performed well, but the results were still below the best results of VGG-16 and emphasized that the VGG-16 architecture, especially with a data separation of 90%:10%, is the most reliable CNN architecture for accurately classifying Toraja carvings.

Related SDGs Targets

Full Author List

Azis, Huzain; Nasrullah, Departement of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia; Abdullah Munaisyah; Ismail, Suriana, Malaysian Institute of Information Technology, Universiti Kuala Lumpur, Kuala Lumpur, Malaysia; Purnawansyah; Syafie, Lukman, , Departement of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

The use of Unmanned Aerial Vehicles (UAVs) in aerial imaging is expanding rapidly, particularly in traffic monitoring and intelligent transportation systems. Detecting small and occluded vehicles in aerial images poses significant challenges due to varying resolutions and obstructions like buildings or trees. This study seeks to enhance vehicle detection accuracy by improving You Only Look Once (YOLO) models, with a focus on small and occluded object detection. Utilizing the COWC-M dataset and advanced data augmentation techniques such as Mosaic Augmentation, this research evaluates multiple YOLO variants. The YOLOv8-L model achieved the highest mAP50 score of 0.9899, demonstrating superior detection accuracy for small objects. Additionally, the YOLOv10-L model outperformed others with the best mAP50–95 score of 0.8715, indicating strong results across different intersection-over-union (IoU) ranges. Compared to YOLO-RTUAV, which achieved an mAP50 of 0.9353, the newer YOLO models provide significant improvements in both precision and recall. These findings contribute to the development of highly efficient, real-time vehicle detection systems suitable for large-scale surveillance applications in complex environments.

Related SDGs Targets

Full Author List

Syafie, Lukman, Universiti Kuala Lumpur, 50250 Kuala Lumpur, Malaysia; Rismayanti, Universitas Negeri Malang, Kota Malang, Jawa Timur 65145, Indonesia

Abstract

Benign Prostatic Hyperplasia (BPH) is a prevalent non-cancerous enlargement of the prostate gland in aging men, often requiring early diagnosis to prevent urinary complications and improve patient outcomes. Traditional diagnostic procedures are limited by subjectivity and accessibility, especially in under-resourced regions. This study proposes an automated diagnostic approach using a deep learning model based on DenseNet121 to classify RGB prostate images into BPH and normal categories. A region-specific dataset consisting of 176 labeled RGB images, collected from a clinical facility in Bangladesh, was used to train and evaluate the model. Pre-processingincluded image resizing, normalization, and data augmentation to enhance generalization. Transfer learning was employed to fine-tune the model, which was trained over 10 epochs using the Adam optimizer and cross-entropy loss. The model achieved a best validation accuracy of 94.12%, with a recall of 72.2% for BPH detection, demonstrating its ability to identify pathological patterns in simple imaging modalities. Despite challenges such as dataset size and imbalance, the findings indicate that RGB image-based deep learning models can support clinical diagnosis of BPH in low-resource settings. This work contributes a lightweight, accessible solution for prostate disease screening and provides a foundation for future research on scalable AI-assisted diagnostics

Related SDGs Targets

Full Author List

Herman; Putri, An’nisa Pratama, Universitas Muslim Indonesia,Makassar, Sulawesi Selatan 90234,Indonesia; Noor, MegatNorulazmi Megat Mohamed, MIIT University Kuala Lumpur, 50250 Kuala Lumpur,Malaysia; Darwis, Herdianti; Hayati, Lilis Nur; Irawati; As'ad, Ihwana, Universitas Muslim Indonesia,Makassar, Sulawesi Selatan 90234,Indonesia

Abstract

Introduction: Passura', or Toraja carvings, are an essential element of the cultural heritage of the Toraja people in Indonesia. These carvings feature complex motifs rooted in nature, folklore, and spiritual symbolism. This study aims to evaluate the efficacy of four Convolutional Neural Network (CNN) architectures—VGG-16, DenseNet121, ResNet50V2, and EfficientNetB0—in classifying seven traditional Toraja carving motifs. Methods: A dataset of 700 images was collected and categorized into seven motif classes. The dataset was split into 80% for training and 20% for validation. Each CNN model was trained for 25 epochs with standard pre-processing, including resizing to 224×224 and normalization. Performance evaluation was conducted based on validation accuracy and confusion matrix analysis to assess classification precision and model overfitting. Results: EfficientNetB0 achieved the highest validation accuracy of 98%, although signs of overfitting were observed. ResNet50V2 followed closely with a validation accuracy of 95.33% and demonstrated the most balanced classification results across all motif categories. VGG-16 and DenseNet121 achieved 94.67% and 81.82%, respectively. Confusion matrix analysis confirmed the robustness of ResNet50V2 in correctly identifying complex patterns. Conclusions: The findings indicate that ResNet50V2 provides a reliable balance between accuracy and generalizability for classifying Toraja carvings, making it suitable for digital preservation of cultural heritage. EfficientNetB0, while achieving higher accuracy, may require additional regularization to avoid overfitting. This study contributes to the development of AI-driven cultural documentation and suggests future research with larger and more diverse datasets to improve model robustness

Related SDGs Targets

Full Author List

Syafie, Lukman; Azis, Huzain; Admojo, Fadhilah Tangguh, Universiti Kuala Lumpur, 50250 Kuala Lumpur, Malaysia

Abstract

Introduction: The preservation of Javanese script as part of Indonesia’s cultural heritage is increasingly urgent in the digital era, especially due to declining literacy among younger generations. This study aims to explore the effectiveness of YOLOv8, an advanced object detection algorithm, for recognizing handwritten Javanese numerals to support efforts in cultural digitization and education. Methods: A dataset of 2,790 handwritten Javanese numerals (0–9) was collected from 93 respondents. Each numeral was manually annotated using bounding boxes via the MakeSense.ai platform. The YOLOv8 model was trained using 80% of the data and validated on the remaining 20%. Training was performed in the PyTorch framework with data augmentation techniques to increase robustness. Model performance was evaluated using precision, recall, F1-score, and mean Average Precision (mAP), along with visualization through confidence curves and confusion matrices. Results: The model achieved a high validation precision of 88.3%, recall of 89.1%, and mAP of 0.88 at IoU 0.90. F1-score peaked at a confidence threshold of 0.89, while certain numerals like 'six' and 'nine' achieved near-perfect detection. Visualizations confirmed the model’s ability to accurately classify and localize characters in both training and unseen data. Minor misclassifications occurred between visually similar numerals. Conclusions: YOLOv8 demonstrates high effectiveness in recognizing handwritten Javanese numerals and holds significant potential for digital heritage preservation. Future work should focus on expanding the dataset, improving generalization under varied conditions, and integrating this model into educational tools and augmented reality applications for interactive learning.

Related SDGs Targets

Full Author List

Nanda, Sultan Aziz Syaifullah, rogram Studi Teknik Informatika,Universitas Muslim Indonesia, Jalan Urip Sumoharjo, Makassar 90231, Indonesia; Azis, Huzain, Universiti Kuala Lumpur, 50250 Kuala Lumpur, Malaysia

Abstract

This research aims to design and develop a power control and monitoring device based on NodeMCU and the PZEM-004T module as the central controller and monitoring unit, enabling users to remotely control and monitor household lamp power consumption. The method used in this study is an experimental approach to evaluate the performance of the developed lamp power control and monitoring device. The NodeMCU ESP8266 microcontroller is employed to execute program instructions and enable remote control, while the PZEM-004T module functions as the power consumption meter. Through serial communication, it provides real-time data on power (Watt), energy consumption (kWh), and cost (IDR) to a smartphone via the Thinger.io web platform. Users can manage the ON-OFF status of home lighting and monitor power usage in real-time. The testing results indicate that the developed prototype successfully supports real-time control and monitoring with an average power measurement error of only 0.10%.

Related SDGs Targets

Full Author List

Latip, Abdul, Institute of Advanced Informatics and Computing, Tasikmalaya 46115, Indonesia; Azis, Huzain, Malaysian Institute of Information Technology, Universiti Kuala Lumpur, Malaysia; Himawan, Hidayatulloh, Department of Informatics, UPN Veteran Yogyakarta, Indonesia, Faculty of Information and Communication Technology, Universiti Teknikal Malaysia, Malaysia; Kurnia, Dian Ade, Department of Informatics Management, STMIK IKMI Cirebon, Indonesia, Faculty of Information and Communication Technology, Universiti Teknikal Malaysia, Malaysia

Abstract

Cirrhosis is a chronic liver disease with significant health implications, responsible for 56,585 deaths annually, and ranking as the 9th leading cause of mortality worldwide. Early detection is crucial for effective treatment and better patient outcomes, as cirrhosis can progress to irreversible damage if not addressed in its initial stages. This research focuses on developing an advanced, integrated method for detecting cirrhosis by employing a combination of Synthetic Minority Over-sampling Technique (SMOTE) and machine learning models, specifically Gradient Boosting and XGBoost. The use of SMOTE is critical in this study as it addresses class imbalance in the dataset, which is a common challenge in medical diagnosis problems, especially when dealing with rare or minority conditions like cirrhosis. Class imbalance can lead to biased models that perform poorly on the minority class, which, in this case, could mean missing crucial cirrhosis diagnoses. SMOTE oversamples the minority class to ensure a more balanced dataset, which improves the model's ability to detect cirrhosis accurately. The research further includes a performance comparison between two powerful machine learning algorithms: Gradient Boosting and XGBoost. Gradient Boosting is known for its ability to optimize the model by focusing on misclassified instances in a sequential manner, while XGBoost, an advanced version of Gradient Boosting, is renowned for its speed and efficiency due to parallel processing and advanced regularization techniques.

Related SDGs Targets

Full Author List

Suzanti, Ika Oktavia; Kamil, Fajrul Ihsan; Rochman, Eka Mala Sari; Rachman, Fika Hastarita; Solihin, Firdaus, Informatics Engineering Department, University of Trunojoyo Madura, Bangkalan, Indonesia; Huzain Azis, Malaysian Institute of Information Technology (MIIT), Universiti Kuala Lumpur, Kuala Lumpur, Malaysia; Suni, Alfa Faridh, Computer Science Department, Newcastle University, United Kingdom

Abstract

Hidden paradise is a term that aptly describes the island of Madura, which offers diverse tourism potential. Through the Google Maps application, tourists can access sentiment-based information about various attractions in Madura, serving both as a reference before visiting and as evaluation material for the local government. The Multinomial Naïve Bayes method is used for text classification due to its simplicity and effectiveness in handling text mining tasks. The sentiment classification is divided into three categories: positive, negative, and mixed. Initial analysis revealed an imbalance in sentiment data, with most reviews being positive. To address this, sampling techniques—both oversampling and undersampling—were applied to achieve a more balanced data distribution. Additionally, the Adaptive Boosting ensemble method was used to enhance the accuracy of the Multinomial Naïve Bayes model. The dataset was split into training and testing sets using ratios of 60:40, 70:30, and 80:20 to evaluate the model’s stability and reliability. The results showed that the highest F1-score, 84.1%, was achieved using the Multinomial Naïve Bayes method with Adaptive Boosting, which outperformed the model without boosting, which had an accuracy of 76%.

Related SDGs Targets

Full Author List

Herman; Akbar, Muhammad; Hayati, Lilis Nur, Information System, Faculty Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia; Nasir, Haidawati, Information Technology, MIIT, University Kuala Lumpur, Kuala Lumpur, Malaysia; Herdianti; Azis, Huzain, Informatic, Faculty Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia

Abstract

The complexity of the motifs and large number of different patterns make the classification of Toraja carvings challenging. The objective of this study is to develop a Convolutional Neural Network automatic classification model using a comparative analysis of the performance of three ResNet architectures. Data augmentation techniques were used to enrich the diversity of the training samples and improve the robustness of the model. The experimental results showed that ResNet101V2 had the highest validation accuracy, which was greater than 97%, followed by ResNet50V2 with more than 96%, and finally, ResNet152V2 with more than 94.74%. These test results indicate that the ResNet101V2 architecture has a better classification performance for complex motifs, with a good balance between precision and recall. However, the confusion matrix and per-class performance metrics indicated that motifs with high similarity, such as Paqdon-Bolu and Paqtedong, remained challenging. This study demonstrated that deeper CNN architectures and data augmentation techniques are effective in improving the classification accuracy of complex carving patterns. Further research should explore hybrid or advanced augmentation methods to improve the overall robustness and accuracy of the model.

Related SDGs Targets

Full Author List

Darwis, Herdianti; Purnawansyah; Umalekhoa, Alfi Syahrin; Adnan, Adam; Umar, Fitriyani; AR, Muh. Aqil Fajar, Universitas Muslim Indonesia, Jl. Urip Sumoharjo KM 5, 90231, Makassar, Indonesia; Salim, Yulita, Universiti Kuala Lumpur, 1016 Jl. Sultan Ismail, Bandar Wawasan, 50250, Kuala Lumpur, Malaysia; Raja, Roesman Ridwan, Kyushu Institute of Technology, 680-4 Kawazu, Iizuka, 820-8502, Fukuoka, Japan

Abstract

This research investigates the application of clustering techniques to network congestion data at Universitas Muslim Indonesia, employing a hybrid metric approach based on packet loss and delay. The study utilized two algorithms, K-Means and K-Medoid, applied in a semi-supervised scenario to group 255,147 network data points into 3, 4, and 5 clusters, considering 10 principal variables. During the pre-processing phase, data cleansing was conducted to address missing values, followed by normalization to standardize the scale of numerical variables, thereby preparing the data for the clustering process. Model validation was performed using four cluster evaluation methods: Gap Statistic, Davies-Bouldin Index, and Elbow Method. The evaluation results indicate that both algorithms were capable of forming valid and reliable clusters. However, the K-Means algorithm demonstrated superior performance compared to K-Medoid, particularly when utilizing three Quality of Service variables: throughput, packet loss, and delay. In this configuration, K-Means yielded more stable clusters, a clearer separation between clusters, and a more structured visualization. Consequently, K-Means is considered more optimal for classifying network congestion levels and presents an effective approach for network data segmentation

Related SDGs Targets

Full Author List

Saly, Intan Novita; Purnawansyah, Universitas Muslim Indonesia, Sulawesi Selatan, Makassar, 90231, Indonesia; Azis, Huzain, Universitas Muslim Indonesia, Sulawesi Selatan, Makassar, 90231, Indonesia, Universiti Kuala Lumpur, Kuala Lumpur, Malaysia, 50250, Malaysia

Abstract

This study aims to evaluate the performance of various classification methods in recognizing handwritten Arabic numerals, particularly the K-Nearest Neighbors (KNN), Gaussian Naive Bayes (GNB), and NU Support Vector Classifier (NU SVC) algorithms. In this study, a dataset of handwritten Arabic numerals consisting of 9,350 samples with 10 different classes was used. The research process involved data collection, data labeling, dividing the dataset into training and testing data, implementing classification algorithms, and performance testing using cross-validation methods. The results showed that NU SVC had more stable performance with accuracy close to KNN, while GNB showed the lowest performance. The conclusion of this study emphasizes that the selection of algorithms and parameter optimization is crucial to improve the accuracy and efficiency of handwriting recognition systems. Support Vector Machine (SVM) based algorithms proved to be superior in handling complex classification tasks compared to GNB. This study provides significant contributions to the field of handwriting recognition, particularly in the context of Arabic numeral handwriting, and can serve as a reference for developers of optical character recognition (OCR) systems in the future. Future research is recommended to increase the variety of datasets and further explore parameter optimization and data preprocessing techniques to improve system accuracy.

Related SDGs Targets

Full Author List

Bustam, Faida Daeng; Purnawansyah, Universitas Muslim Indonesia, Sulawesi Selatan, Makassar, 90231, Indonesia; Azis, Huzain, Universitas Muslim Indonesia, Sulawesi Selatan, Makassar, 90231, Indonesia, Universiti Kuala Lumpur, Kuala Lumpur, Malaysia, 50250, Malaysia

Abstract

This research explores the performance of several classification algorithms on handwritten Lontara digits, a script traditionally used by the Bugis and Makassar communities in South Sulawesi, Indonesia. The dataset comprises 10,890-digit samples, contributed by 99 individuals, and is categorized into 10 distinct classes corresponding to the digits 0- 9. The classification methods evaluated in this study include K-Nearest Neighbors (KNN), Gaussian Naive Bayes (GNB), and Nu-Support Vector Classifier (NuSVC). Cross-validation techniques are employed to evaluate the performance of these classifiers using standard metrics such as accuracy, precision, recall, and F1 score. The findings demonstrate varying levels of performance across the algorithms. Notably, GNB achieves the highest recall, indicating its ability to correctly identify positive samples, whereas KNN and NuSVC exhibit moderate effectiveness across other performance metrics. KNN shows potential with its simple yet robust approach to classifying complex datasets, while NuSVC demonstrates a balanced performance, particularly in precision. However, all classifiers face challenges in achieving optimal accuracy, particularly due to the complexity of the handwritten Lontara digits, which exhibit unique and intricate patterns. The study concludes by suggesting that further improvements can be achieved by refining feature extraction techniques and optimizing the classifiers used. Enhancing feature extraction could provide better representations of the Lontara digits, potentially leading to improved classification accuracy. Additionally, algorithm optimization and the exploration of more advanced classification methods could further enhance the overall performance. This research provides a foundation for the development of automated recognition systems for Lontara script, contributing to its preservation and modern use.

Related SDGs Targets

Full Author List

Herman, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia; Nasir, Haidawati; Noor, Megat Norulazmi Megat Mohamed, Computer Engineering Technology, MIIT Universiti Kuala Lumpur, Kuala Lumpur, Malaysia; Hasanuddin, Tasrif; Indra, Dolly, Faculty of Computer Science, Universitas Muslim Indonesia, Makassar, Indonesia; Lumentut, Hence Beedwel, Computer Engineering, STMIK Agamua Wamena, Wamena, Indonesia

Abstract

Toraja is a part of Indonesia located on the island of Sulawesi. One form of Toraja culture is carving art made from wood, bamboo or stone. There are around 70 types of Toraja carving motifs. With so many motifs and some motifs being almost similar to each other, it can make it difficult for the public and tourists to know the name of the motif. This research focuses on classifying Toraja carving images using Convolutional Neural Network (CNN) and measuring the performance of the LeN et-5 architecture. The number of Toraja carving motifs used in this research is 7 and the data is 700, where each motif is represented by 100 data. In testing, 3 explorations were carried out, namely dividing training and validation data, Batch Size values, and Target Size values. Based on test results, the 70:30 data division provides the highest level of accuracy compared to other data divisions. Increasing the Batch Size value has a negative impact on the level of accuracy. On the other hand, increasing the Target Size value has a positive impact on the accuracy value. The best Batch Size and Target Size values in this study were 32 and 256x256 with respective accuracies of 50.67% and 69.33%.

Related SDGs Targets