Indonesian Journal of Electrical Engineering and Computer Science
Not a member yet
9109 research outputs found
Sort by
A hybrid divisive K-means framework for big data–driven poverty analysis in Central Java Province
Get PDFClustering is essential in big data analytics, especially for partitioning high dimensional socioeconomic datasets to support interpretation and policy decisions. While K-Means is widely used for its simplicity and scalability, its strong sensitivity to initial centroid selection often leads to unstable results and slower convergence. Previous hybrid approaches, such as Agglomerative–K-Means, attempted to address this issue by using hierarchical clustering for centroid initialization; however, these methods rely on bottom-up merging, which can produce suboptimal initial partitions and increase computational overhead for larger datasets. To overcome these limitations, this study proposes a hybrid divisive–K-Means (DHC) model that employs top-down hierarchical splitting to generate more coherent initial centroids before refinement with K-Means. Using a multidimensional poverty dataset from Central Java Province provided by the Indonesian Central Bureau of Statistics (BPS), the performance of DHC was evaluated against standard K-Means and Agglomerative–K-Means. The assessment included execution time, convergence iterations, and cluster validity indices (Silhouette, Davies–Bouldin, and Calinski–Harabasz). Experimental results demonstrate that DHC reduces execution time by up to 97% and requires 40% fewer iterations than standard K-Means, while achieving comparable or improved cluster quality (e.g., CH Index increasing from 14.3 to 15.8). These findings indicate that the DHC model offers a more efficient and stable clustering solution, addressing the shortcomings of previous standard K-Means methods and improving performance for large-scale socioeconomic data analysis
Detection of COVID-19 using chest X-rays enhanced by histogram equalization and convolutional neural networks
Get PDFThe persistent global health crisis initiated by the COVID-19 pandemic continues to demand robust and high-throughput diagnostic solutions. While gold-standard methods, such as polymerase chain reaction (PCR) testing, are accurate, their scalability and turnaround time remain limitations in high volume settings. This paper introduces a novel deep learning framework designed for rapid and accurate detection of COVID-19 from chest X-ray (CXR) imagery. Our methodology leverages a convolutional neural network (CNN) architecture, augmented by a crucial pre-processing stage: histogram equalization. This step is vital for enhancing the subtle contrast features inherent in CXR scans, there by significantly improving the quality of the input data and facilitating superior feature extraction by the CNN. The model was trained and rigorously validated on a dedicated dataset. Performance was systematically quantified using a comprehensive confusion matrix, yielding key metrics such as precision and specificity, alongside the receiver operating characteristic (ROC) curve. The achieved results are highly encouraging, demonstrating a classification accuracy of 98.45%. This innovative approach offers a substantial acceleration of the diagnostic process, providing a non-invasive and highly effective complementary tool for clinicians. Ultimately, this advancement has the potential to streamline patient management protocols and alleviate diagnostic pressures on global healthcare infrastructures
Invisible watermarking as an additional forensic feature of e-meterai
Get PDFThe e-meterai is an official digital product of the Indonesian government issued by the Directorate General of Taxation (DGT). Its usage has become increasingly widespread as conventional documentation transitions to digital formats, serving the same function as its printed counterpart. This product features a quick-response code embedded with unique Indonesian codes and offers overt, covert, and forensic features. This study aims to experiment with adding a forensic feature in the form of an invisible watermark. We employed two watermark embedding techniques, discrete Fourier transform (DFT) and scale-invariant feature transform (SIFT), to determine which is more suitable for this application. After embedding the watermark, we also simulate various attacks including gaussian noise, salt and pepper noise, averaging filter, rotation, translation, and speckle noise. For each attack, we calculated with normalized-cross correlation (NCC) values, obtaining 0.863 and 0.976 for the gaussian noise attack, 0.929 and 0.984 for the salt and pepper attack, 0.975 and 0.984 for the averaging filter attack, 0.173 and 0.097 for rotation attacks, 0.172 and 0.032 for translation attack, and 0.972 and 0.996 for speckle noise attack, using DFT and SIFT techniques, respectively
Tool support for LoRaWAN development: a comparative perspective on simulation and emulation
Get PDFThis paper explores the use of various long range wireless area network (LoRaWAN) simulation and emulation tools when designing and evaluating IoT networks. Simulation tools are often popular with researchers because they are less costly and can easily simulate large-scale networks, allowing for easy and faster tests of the scalability of various protocols and behaviors. However, they often lack the unpredictable nature of real deployments. Emulation and cloud-based tools fill this gap, but with their flexibility they provide a more realistic approximation of real-world performance and allow easier interfacing with actual network hardware infrastructure, although they generally incur a higher cost which is often controlled by technical skill level use.
Development of a machine learning model with optuna and ensemble learning to improve performance on multiple datasets
Get PDFMachine learning, a subset of artificial intelligence (AI) is vital for its ability to learn from data and improve system performance. In Indonesia, advancements in ML have significant potential to boost competitiveness and foster sustainable development. However, issues like overfitting and suboptimal parameter settings can hinder model effectiveness. This study aims to improve the classification performance of ML models on various datasets. Advanced techniques like hyperparameter tuning with Optuna and ensemble learning with extreme gradient boosting (XGBoost) are integrated to enhance model performance. The study evaluates the performance of K nearest neighbors (KNN), support vector machine (SVM), and Gaussian naïve Bayes (GNB) algorithms across three datasets: academic records from the Islamic University of Riau (UIR), diabetes data from Kaggle, and Twitter data related to the 2024 elections. The findings reveal that the GNB algorithm outperforms KNN and SVM across all datasets, achieving the highest accuracy, precision, recall, and F1-score. Hyperparameter tuning with Optuna significantly improves model performance, demonstrating the value of systematic optimization. This study highlights the importance of advanced optimization techniques in developing high-performing ML models. The results suggest that robust algorithms like GNB, combined with hyperparameter tuning and ensemble learning, can significantly enhance classification performance
Prediction of permeability via nuclear magnetic resonance logging using convolutional neural networks
Get PDFPermeability is a critical parameter in subsurface fluid flow analysis, reservoir management, hydrocarbon recovery, and carbon dioxide sequestration. Traditional permeability measurement methods involve costly and time-consuming laboratory tests or well-related data. Machine learning (ML), specifically convolutional neural networks (CNN), is proposed as a cost-effective and rapid permeability prediction solution, harnessing interrelationships of input-output variables. In this study, empirical permeability correlation was developed using CNN. Forty nuclear magnetic resonance (NMR) T2 spectrums and 89 logarithmic mean NMR T2 distributions (T2lm) were preprocessed, screened and key spectra were identified using the principal component analysis (PCA). To develop the correlations, a custom-designed CNN architecture was employed to leverage the spatial patterns and intricate relationships embedded in the NMR data. The model was trained and validated rigorously using k-fold cross validation scheme to ensure robustness and generalization. Performance metrics like R-squared (R2), root mean squared error (RMSE), mean absolute error (MAE), standard deviation (SD), absolute deviation (AD), average absolute deviation (AAD), average absolute percentage relative error (AAPRE), and maximum error (Emax) were deployed to evaluate the model’s accuracy and ability to predict permeability values accurately. Among the folds considered, the fold 1 emerged as the best-performing model with the highest R2 value of 0.9544. This CNN-based correlation outperformed conventional and other AI-based models in terms of R2, Emax, AD, AAD, AAPRE, among other metrics. Overall, the study demonstrates the effectiveness of CNN in predicting permeability, offering a superior alternative to costly and limited traditional methods, with fold 1 showing the most promising results
Dynamic behavior of induction machines in ATP-EMTP with space harmonics
Get PDFThis work develops a space-vector model of a squirrel-cage induction machine that incorporates the effects of spatial harmonics arising from the winding distribution. The modeling approach includes the first, fifth, and seventh spatial harmonics, which are the components with the greatest influence on the machine’s magnetic field. Simulation results highlight the impact of these harmonics on the stator and rotor currents, the electromagnetic torque, and the machine’s speed. To build the model, the voltage behind reactances (VBR) technique is employed, enabling a hybrid strategy that combines circuit-based modeling tools—such as ATP-EMTP—with computational programming in models to complement the solution of the differential equations governing the behavior of the electromechanical system. This methodology effectively transforms the induction machine into a dynamic Thevenin-equivalent circuit for each phase of the converter. ` This study provides a useful framework for evaluating how space harmonics affect the performance and operating characteristics of induction machines. The models were implemented using the ATP-EMTP software and its graphical interface, ATPDraw
Intelligent dust monitoring and cleaning optimization on photovoltaic panels
Get PDFDust deposition on photovoltaic (PV) panels is a significant operational issue, often leading to power losses exceeding 15–30% in regions with high airborne particle concentrations. Although numerous studies have investigated either visual detection of dust or analytical estimation of performance loss, most approaches focus on a single task and provide limited practical insight for real-time maintenance. This work introduces a dual-task deep learning framework that simultaneously classifies dust severity and predicts the corresponding power loss from panel images. Five recent architectures vision transformer (ViT), swin transformer, GhostNet, DenseNet, and MobileNetV2 are employed as backbone feature extractors, with extracted embeddings processed by a multi-head multi-layer perceptron (MLP) combining shared representation learning with separate classification and regression outputs. The system is trained and evaluated on a real-world dataset of PV panels, and performance is assessed using accuracy and mean absolute error. DenseNet achieves the highest accuracy (94%) and lowest prediction error, while lightweight convolutional neural network (CNN) backbones demonstrate the best balance between precision and computational efficiency. By integrating hybrid processing and dual predictive capability, the proposed method offers a more comprehensive and deployable solution for automated PV monitoring compared to existing single-output approaches
Comparative analysis of fractional-order sliding mode and pole placement control for robotic manipulator
Get PDFFractional-order sliding mode control (FOSMC) is benchmarked against pole placement control (PPC) on a nonlinear two-link manipulator subjected to identical trajectories and 10 N·m square disturbances. Quantitative head-to-head evidence against industrial PPC is scarce, leaving engineers uncertain when fractional designs justify their added complexity. We derive the plant via Lagrange dynamics, implement both controllers in Python, and evaluate tracking and torque effort using SciPy-based simulations. Under the adopted fractional derivative approximation, FOSMC attains RMSEs of 0.458 rad (q1) and 0.453 rad (q2) whereas PPC limits the errors to 0.365 rad and 0.337 rad. The fractional design, however, requires lower mean torques of 69.2/29.0 N·m compared to PPC’s 86.1/41.4 N·m, exposing a precision–energy trade-off that now favours PPC on accuracy and FOSMC on actuation effort. The benchmark delivers deployable evidence that fractional sliding surfaces shift torque demand even when their tracking performance lags, and it motivates hardware-in-the-loop validation to close the identified accuracy gap
Deep-fuzzy personalisation framework for robot-assisted learning for children with autism
Get PDFResearch exploring the efficacy of robots in autism therapy has predominantly relied on the Wizard-of-Oz method, where robots execute predetermined behaviours. However, this approach is constrained by its heavy reliance on human intervention. To address this limitation, we introduce a novel deep-fuzzy personalization framework for social robots to enhance adaptability in interactions with autistic children. This framework incorporates a deep learning model called singleshot emotion detector (SED) with a mean average precision of 93% and a fuzzy-based engagement prediction engine, utilizing factors such as scores, IQ levels, and task complexity to estimate the engagement of autistic children during robot interactions. Implemented on the humanoid robot RoCA, our study assesses the impact of this personalization approach on learning outcomes in interactions with Ghanaian autistic children. Statistical analysis, specifically Mann Whitney tests (U=3.0, P=0.012), demonstrates the significant improvement in learning gains associated with RoCA's adoption of the deep fuzzy approach