Indonesian Journal of Electrical Engineering and Computer Science
Not a member yet
9109 research outputs found
Sort by
Word embedding and imbalanced learning impact on Indonesian Quran ontology population
Get PDFThis research addresses limitations in Quranic instance classification, exceptionally high dimensionality, lack of semantic relationships in the term frequency-inverse document frequency (TF-IDF) technique, and imbalanced data distribution, which reduce prediction accuracy for minority classes. This study investigates the impact of word embedding and imbalance learning techniques on instance classification frameworks using Indonesian Quran translation and Tafsir datasets to handle previous research limitations. Four classification frameworks were built and evaluated using accuracy and hamming loss metrics. The results show that the synthetic minority oversampling technique (SMOTE) technique, TF-IDF model, and logistic regression classifier provide the best accuracy results of 62.74% and a hamming loss score of 0.3726 on the Quraish Shihab Tafsir dataset. This is better than the performance of previous classifiers backpropagation neural network (BPNN) and support vector machine (SVM) used in the previous framework, with accuracies of 59.91% and 62.26%, respectively. Logistic regression can also provide the best classification results with an accuracy of 67.92% and a hamming loss of 0.3208 using the previous framework. These results are better than the performance of the previous classifiers BPNN and SVM used in the previous framework, with accuracies of 62.26% and 66.98%, respectively. TF-IDF feature extraction outperforms word2vec in instance classification results due to its superior support under limited dataset conditions
Analysis of solid oxide fuel cell systems for off-grid energy production
Get PDFThis work presents a simulation study of a 50-kW solid oxide fuel cell (SOFC) power supply system that provides electricity to residential users. Indeed, many decentralized applications rely on renewable energy sources not connected to the primary power grid. Moreover, fuel cell modelling and simulation are critical for promoting renewable energy as they eliminate the need for physical prototypes, saving time and money. We have also developed a reliable model for simulating self-contained SOFC fuel cells. The elaborated model includes the kinetics of electrochemical processes and accounts for voltage losses in SOFCs. Our fuel cells produce the necessary electrical current to charge the device. Also, our system has fuel cells, a DC/DC converter, and an inverter with LCL filters. These components connect the fuel cell system to other power electronics and the electrical load. Furthermore, a mathematical model of a dual inverter system describes its control method, including the proportional and integral parameters in the voltage and current loops has been derived. The proposed model and system could be helpful for a standalone load
Development and evaluation of a generalized ontology framework for software requirement specification
Get PDFThis paper presents an ontology developed to address challenges such as com munication gaps, risks of errors, and inconsistencies during the manual process of creating software requirement specifications (SRS). The proposed ontology offers a systematic and formal depiction of the requirements, enhancing consis tency and communication among stakeholders. The ontology has been devel oped from the software requirements documents to facilitate the development process. This paper discusses the process of creating the ontology and demon strates using Pellet Reasoner for inference and Prot´eg´e for ontology construction to save and reuse information. The ontology seems to be efficient in manag ing complex software projects, enabling accurate requirement retrieval through SPARQL queries. This study emphasizes how incorporating ontologies into re quirement engineering can significantly enhance the quality and reliability of SRS
Enhancing evaluation practices for Islamic inheritance calculation systems: toward a standardized benchmark
Get PDFAccurate estate distribution is a critical aspect in Islamic law, governed by complex rules that require precise inheritance calculations. Although numerous computerized inheritance calculation systems have been developed, their reliability remains questionable due to inadequate evaluation and unclear criteria for test case selection. This study addresses this gap by introducing a structured evaluation methodology to rigorously assess the functionalities of inheritance calculation systems. A new benchmark comprising 50 test cases was developed by reviewing the functionality of existing systems, collecting prior test cases and identifying coverage gaps through a detailed gap analysis. These benchmark cases were then used to assess the performance of leading online inheritance calculators, comparing their results to expert-validated solutions. Results revealed a significant drop in performance for calculators previously reported to achieve near-perfect accuracy, with scores declining to 68% and 58% compared to earlier reports of 100% and 90%. This demonstrates the effectiveness of the proposed test cases in exposing limitations within current systems. In contrast, the Almwareeth calculator, which had not been previously evaluated, demonstrated the highest accuracy (86%) and was able to handle a wider range of cases. This study lays a critical foundation for advancing the evaluation standards of Islamic inheritance calculation systems, thereby enhancing their reliability in real-world applications
Interpretable federated deep learning models for predicting gait dynamics in biomechanics
Get PDFAccurate prediction of human joint angle dynamics and reliable gait classifica tion are essential for applications in rehabilitation, biomechanics, and clinical monitoring. Traditional machine learning (ML) models trained on centralized data raise concerns about privacy, scalability, and transparency. This study proposes a federated deep learning (DL) framework that integrates privacy preserving model training with interpretable predictions. Specifically, a gated recurrent unit- deep neural network (GRU-DNN) hybrid model is developed for regression of joint angles, while a Long short-term memory- convolutional neural network (LSTM-CNN) hybrid model is designed for binary and multi class gait classification. The framework is deployed using the federated av eraging (FedAvg) algorithm across simulated clients, with each client training locally on its data. To enhance interpretability, the local interpretable model agnostic explanations (LIME) algorithm is integrated at the client level to gener ate human-understandable explanations for model predictions. The experimen tal results demonstrate significant improvements, including a reduction in global mean squared error (GMSE) from 56.16 to 3.31 and an increase in R-squared score from 0.80 to 0.99 for regression, along with classification accuracies of 0.97 (binary) and 0.94 (multi-class). This scalable, privacy-preserving frame work bridges the gap between accuracy and transparency, offering impactful applications in biomechanics, healthcare, and personalized medicine
Optimizing clustering efficiency with weighted k-means: a machine learning-driven approach for enhanced accuracy and scalability
Get PDFData analysis unlocks the hidden, latent patterns and structures within datasets. Clustering algorithms, the cornerstone of any data analysis, are usually challenged by high-dimensionality, complexity, or large-scale data. This research proposes a hybrid model that merges neural networks and clustering techniques to handle these problems. Neural networks are used for feature extraction and dimensionality reduction; raw data will be transformed into a robust, low-dimensional representation. With these refined features, the performance of clustering algorithms improves in terms of scalability, efficiency, and accuracy. The proposed model is tested on diversified datasets such as the wisconsin breast cancer dataset (WBCD), GEO Dataset, and image and text data benchmarks for which substantial improvements in clustering metrics such as silhouette score, purity, and computational efficiency are reported. The results demonstrate the efficacy of the hybrid approach in optimizing clustering applications across domains, such as bioinformatics, health care, and image analysis
Development and integration of a privacy computing gateway for enhanced interoperability
Get PDFA new design of privacy computing gateway stands as the solution to secure efficient interoperability between heterogeneous platforms. The growing importance of data privacy, along with rising collaborative data analysis operations, creates an immediate need for standardized privacy-preserving frameworks that are adaptable to diverse situations. A three-layered architecture consisting of application protocol and communication layers receives support from an Adaptation mechanism designed for compatibility between separate privacy computing systems. Testing of the framework uses standard machine learning methods together with horizontal and vertical federated learning using diverse data quantities and feature distribution patterns. The gateway achieves satisfactory model performance and protects data privacy integrity in combination with platform interoperability. area under the curve (AUC) along with F1 score metrics, proves that the proposed system reaches performance equivalence with centralized models when operating within privacy-limited environments. The research introduces an effective solution for securing cross-platform data sharing that will enable secure inter-sector collaboration in finance, healthcare, and government applications
Predicting staple crop yields under climate variability using multiple regression techniques
Get PDFGlobal food systems rely on staple crops—rice, wheat, maize, potato, soybean, and sugarcane, which are vital in Asia, where production is high. However, climate change threatens crop yields, potentially increasing hunger and malnutrition. Yield variability due to climate factors like rainfall and temperature underscores the need for accurate crop yield predictions. This paper analyzed the relationships between staple crop yields, climate variables, and pesticide usage. It aimed to develop a predictive model for crop yields in Asia using multiple regression techniques in Google Colab. The model was evaluated using a hybrid set of metrics like mean absolute error (MAE), root mean squared error (RMSE), and R² score. Findings revealed that reliable yield predictions are achievable despite weak linear relationships among variables. The extreme gradient boosting (XGBoost) achieves the highest R² score of 0.958367, which indicates superior predictive performance for staple crop yield forecasting due to its lower overall error rates and greater consistency in performance. This highlights the effectiveness of ensemble methods like XGBoost in capturing complex crop yield patterns. Despite newer machine learning (ML) techniques, these models remain recommended for similar tasks due to their robust performance
A novel approach for detection of cracks in painting and concrete surface images using CNN models
Get PDFDiscovering the beginnings of historical artworks takes one on an amazing voyage across space and time. People all around the world have been captivated by India's rich cultural heritage throughout its history, and ancient paintings have always been a very important part of it. Over the period of time, these ancient paintings can get cracks on it due to many factors. This research introduces an automated image classification system where the cracks on the paintings as well as the concrete surface will get detected. Detecting cracks on the concrete surface is important because the longevity and upkeep of concrete structures rely on the prompt identification and treatment of cracks, which can weaken the structure and necessitate expensive repairs. In this study, we focus on image classification using general convolution neural network (CNN), Inception V3, VGG-16, and ResNet-50 models of CNN. These models are trained and validated separately on two different datasets of paintings and concrete surfaces. Inception V3 and VGG-16 models achieve high accuracy, respectively in painting and concrete datasets in comparison with general CNN and ResNet-50 models
Partitioning hazy images using interactive active contour models
Get PDFImage partitioning, also known as image segmentation, is a process that involves dividing an image into distinct and meaningful segments. Recently, an interactive active contour model (ACM) namely the Gaussian regularization selective segmentation (GRSS) was designed to handle images with intensity inhomogeneity effectively. However, the GRSS model shows limited performance when applied to hazy images, which often results in incomplete detection and inaccurate extraction of the target object. This study reformulates the GRSS model by integrating the simple dark channel prior (SimpleDCP) dehazing technique, producing a modified model referred to as GRSS with SimpleDCP (GRSSD). The model is derived and implemented in MATLAB software. Experimental results show that the GRSSD model achieves improved segmentation accuracy (ACU) compared with the GRSS model. On average, the ACU improved by 1.8%, while the error metric (EM) decreased from 0.053 to 0.036, representing a reduction of about 32%. The Dice and Jaccard indices improved by approximately 2.6% and 4.9%, respectively. Although the computation time increased, the enhancement in segmentation ACU demonstrates the benefit of incorporating a dehazing process into the variational formulation. The proposed GRSSD model can be extended to color and three-dimensional image segmentation in future work