Indonesian Journal of Electrical Engineering and Informatics (IJEEI)
Not a member yet
776 research outputs found
Sort by
Optimization of Cement Distribution Route Based on Hybrid Genetic-Firefly Algorithm (GAFA)
Full text linkThis study focuses on optimizing the cement distribution route to improve efficiency, reduce costs, and minimize environmental impacts. A hybrid Genetic-Firefly Algorithm (GAFA) approach, integrating the Genetic Algorithm (GA) and the Firefly Algorithm (FA), is developed to solve the complex problem of determining the optimal distribution route to ensure timely, efficient, and sustainable delivery. The Data from PT Semen Baturaja includes three factory locations and 128 distributor points. Various parameter configurations are tested, including population size, mutation probability, total execution time, average execution time, standard deviation of execution time, best factory, and best distance to provide their impact on algorithm performance. The empirical results show that the optimal configuration produces the lowest total distance of 205.14 kilometers and high executiontime efficiency. The best route covers 128 strategic distribution points in the Sumatra region. These results prove the effectiveness of the hybrid GAFA algorithm in optimizing cement distribution routes, contributing significantly to operational efficiency and transportation cost savings. Thus, this approach offers a practical, efficient solution for optimizing cement distribution routes in the manufacturing industry
Transfer Learning for Detecting Alzheimer’s Disease in Brain Using Magnetic Resonance Images
Full text linkAlzheimer’s Disease (AD) is one of the most concerning diseases because the patients show very few symptoms at the earlier stages. Dementia is very common in patients who have suffered brain damage or those who have suffered from psychotic trauma. Patients who have a lot of age suffer the most from it. Magnetic resonance imaging (MRI) is widely used to clinically treat patients with Alzheimer’s. Currently, there is no known remedy for the disease. We can only identify and try to give the proper medications to give some relief to patients. In this study, we have collected MRI data from patients with 4 different stages of Alzheimer’s. The purpose of this paper is to build a model to securely detect these stages for the betterment of medical science. We implemented a transfer learning method with state-of-the-art models such as ResNet50, DenseNet121, and VGG19. We proposed our method with these models which have pre-trained weights of “ImageNet”. The layers that we added are our novelty. We were able to achieve 97.70% accuracy on our best pre-trained model with an F1 score of 97% and a precision of 97% on our test data
Developing a Prototype for Enhancing Data Security in LoRaBased Theft Detection Systems Using ASCON-128 Encryption
Full text linkAsset protection is crucial for organizations to prevent theft. This study presents a LoRa-based theft detection prototype enhanced with ASCON-128 encryption for secure data transmission. The system consists of a transmitter attached to assets and a receiver in a monitoring room, featuring a web-based digital map for real-time tracking. ASCON-128, a NIST-standard lightweight encryption algorithm, ensures data confidentiality and integrity against ManIn-The-Middle (MITM) attacks. The system was evaluated based on transmission speed, power consumption, and security performance. Results indicate that ASCON-128 integration reduces data transmission speed by 42.7% in Line-of-Sight (LOS) and 45.35% in Non-Line-of-Sight (NLOS) conditions. Power consumption increased by 2.7% in standby mode and 12.85% under simulated attack scenarios. Despite these trade-offs, encryption provides significant security benefits with acceptable resource overhead, making it a viable solution for LoRa-based asset tracking and theft detection
Angularly Stable, Transparent and Flexible Modified Octagonal Shaped Frequency Selective Surface (FSS) for Sub-GHz 5G Applications
Full text linkThis paper offers a newly size-reduced Frequency Selective Surface (FSS) featuring band-stop behavior at 4.2 GHz. This developed FSS includes a mushroom-shaped arm with an octagonal patch. The patch is extensively adjusted by incorporating further mushroom-shaped arms, leading to a lower resonance and wider bandwidth. The designed FSS is made up of just a 11 × 11 mm unit cell on a flexible acrylic substrate that is 1 mm thick. The proposed FSS had a 1 GHz bandwidth with a centre resonance frequency of 4.2 GHz. Due to the distinct polarization behavior of this FSS, the Transverse Magnetic (TM) and Transverse Electric (TE) modes are unique and have a steady angular property up to 45º. Measurements of S-parameters for TE and TM polarizations have been validated experimentally over the 2–8 GHz frequency spectrum for both normal and oblique incident angles up to 45°. Excellent agreement between the measured and simulated data is demonstrated, verifying the FSS performance with a frequency variation of less than 3% and preserving constant band-stop properties across all measured orientations. It may be appropriate for incorporation into applicable clothes in a variety of areas because of its simplicity and ease of fabrication
Quantifying Drought Using Machine Learning Models with SPEI indices and Weather Data
Full text linkDrought prediction is crucial for effective water resource management, particularly in regions prone to frequent droughts, such as Rajshahi, Bangladesh. This study presents a novel approach to quantifying and predicting drought conditions in Rajshahi, Bangladesh, utilizing machine learning models with the Standardized Precipitation Evapotranspiration Index (SPEI) as drought indices. We utilized monthly meteorological data (temperature, precipitation, humidity, wind speed, number of sunshine hours, cloud cover, potential evapotranspiration, and the climatic water balance) from 1965 to 2022. To train machine learning models, SPEI drought indicators were numerically encoded and classified into categorical drought situations. To forecast drought conditions in the Rajshahi region, we tested a variety of individual classification and regression algorithms, including Gradient Boosting, XGBoost, Multi-Layer Perceptron (MLP), Random Forest, Logistic Regression, Support Vector Machines, CatBoostClassifier, and Decision Trees. These models performed differently, with accuracy rates ranging from 85% to 88% for classification tests and R² scores from 0.25 to 0.71 for regression tasks. To increase forecast accuracy, we created two hybrid models: the Multi-Model Drought Forecaster and the Drought Anticipation Super Model. The "Multi-Model Drought Forecaster," which combines MLP, Random Forest, Gradient Boosting Classifier, and Decision Tree Classifier, obtained 92% accuracy. The "Drought Anticipation Super Model," incorporating Random Forest, Gradient Boosting, Decision Trees, Support Vector Regression, and CatBoost Classifier, increased the accuracy to 96%. The hybrid model's improved performance demonstrates that it can give more accurate and reliable drought forecasts in the Rajshahi region. These findings improve drought management strategies in Bangladesh and other climate-vulnerable areas. This study also created advanced hybrid machine learning models for drought forecasting in Rajshahi, Bangladesh, with the help of 58 years of meteorological data from 1965 to 2022 and SPEI indices. The “Multi-Model Drought Forecaster” is 92% accurate by utilizing MLP, Random Forest, Gradient Boosting, and Decision Trees. The “Drought Anticipation Super Model” is 96% accurate by adding Support Vector Regression and CatBoost Classifier to provide a better drought forecast to manage water resources effectively
A Survey on Blockchain-Based Routing in Communication Networks
Full text linkRouting in communication networks involves the transmission of packets among network nodes by making routing decisions constructed upon diverse protocols that can depend on various metrics. Blockchain systems made up of concatenated blocks inherently preserve the faithfulness, assure the non-deniability, and assure the obscured individuality of their transactions/blocks through the incorporation of distributed unanimity mechanisms and cryptographic techniques. In the present literature, it clearly lacks a review manuscript on the broad scope of blockchain-based routing; thus, we fill that gap by studying Blockchain-based Routing (BBR) under numerous routing techniques, identifying the concept under 5 divisions, and then in-depth scrutinizing the reviewed work based on blockchain-correlated, routing-correlated, and network-correlated characteristics. We collected a premature sample of 83 articles by cherry-picking articles for qualification requirements explored in scientific databases, employing an in-depth and extended quality assessment approach. As per the appraisal, BBR improves the overall routing performance and security through the storage of routing decisions and updates securely, automatic routing with the aid of smart contracts, providing authentication for secure routing, providing reputation-based routing, and blockchain-based onion routing. In-depth scrutinization reveals that 45.5% of BBR frameworks utilize blockchain for storing routing decisions and updates; 93.2% employ linear blockchain architecture; 20.5% employ proof-of-work consensus; 100% dynamic routing; 72.8% decentralized routing; 93.2% single path routing; 86.4% table-based; and 20.5% are designed for IoT networks. Finally, we disclose the possibilities and impediments of the idea of BBR, identify review gaps, and then render proposals to conquer them
VENTRICULAR TACHYCARDIA PREDICTION THROUGH DEEP LEARNING: ENHANCING CARDIAC MONITORING
Full text linkVentricular Tachycardia (VT) is a life threatening arrhythmia that needs to be detected early and correctly to avoid cardiac arrest. In this paper, the authors hypothesise a hybrid deep learning model based on WaveNet, Swin Transformer, and MISH activation function to make powerful predictions of VTs on the basis of ECG signals in the MIT-BIH Malignant Ventricular Arrhythmia Database (VFDB). The preprocessing pipeline will consist of wavelet-based denoising, min-max normalization and HRV feature. WaveNet has the ability to capture short-term temporal variations whilst the Swin Transformer considers global relations via hierarchical attention. The suggested approach has excellent performance over baseline models, having accuracy, precision, recall, and F1-score of 97.57, 96.89, 97.42, and 97.15, respectively. The improved capability of the model to detect VT with a low number of false negative results shows that the model could be used in realtime cardiac monitoring and clinical decision support. The next steps to be considered in the future research will be the model optimization of wearable devices and testing on multi-center ECG data
Early Mental Health Detection with Machine Learning : A Practical Approach to Model Development and Implementation
Full text linkAcademic pressures, lifestyle changes, and socio-economic factors significantly impact mental health, a critical determinant of academic success and well-being. Early detection and intervention are crucial to mitigate severe outcomes like academic underperformance and suicidal tendencies. Leveraging tools like the DASS-42, this study examines mental health patterns using Support Vector Machine (SVM) models, achieving accuracies of 88% for depression, 71% for stress, and 57% for anxiety. While the model excels in identifying "Normal" cases, its performance for "Mild," "Moderate," and "Severe" cases highlights limitations due to class imbalance and feature representation. The findings reveal that anxiety is the most volatile and severe condition, with peaks in 2018 and 2022, while stress remains manageable and depression moderately stable. Gender and program-specific differences emphasize the need for tailored interventions. Addressing challenges related to data quality, algorithmic transparency, and ethical concerns is essential for real-world applications. This study highlights the potential of machine learning in early detection and intervention for mental health issues. Future research should explore advanced feature engineering techniques and develop more interpretable models to enhance clinical decision-making
Enhancing Intrusion Detection Systems in Cloud Computing Environments: A Hybrid Machine Learning Approach
Full text linkIntrusion Detection Systems (IDS) are essential for maintaining the security of cloud computing environments, which are increasingly targeted by sophisticated cyber-attacks. This paper presents a novel hybrid approach for intrusion detection in cloud environments, combining Random Forest for feature selection, Long Short-Term Memory (LSTM) networks for temporal pattern recognition, and Transformer networks for contextual learning. Evaluated on CICIDS2017 and CSE-CIC-IDS2018 datasets, the proposed approach achieved weighted F1-scores of 97% and 99% respectively, significantly outperforming baseline models. The hybrid model improved accuracy from 95.1% to 98.0% and F1-score from 94.2% to 97.0% compared to LSTM-only approaches. While excelling at detecting common attack patterns such as Distributed Denial of Services (DDoS), challenges remain in identifying rare threats including SQL Injection. This research contributes to cloud security advancement by demonstrating the effectiveness of hybrid machine learning architectures in addressing the unique challenges of intrusion detection in distributed cloud infrastructures
The Effect of Noise on Speaker Identification and Finding a Noise that Improves Accuracy
Full text linkConventional Speaker Identification (SID) systems accurately identify speakers if their speech is noiseless. However, their classification accuracies reduce substantially when speech is corrupted by noise. SID systems would be more practical and applicable if they were more noise-robust. We introduce an SID system that can accurately classify speakers, even when their speech is corrupted by various types of noise at different noise levels. We investigate the impact of noisy training data on the performance of an SID system and the noise that may enhance the performance of an SID system. In this paper, we compare two front-end feature extractors: a cochlea model called the Cascade of Asymmetric Resonators with Fast Acting Compression (CAR-FAC) and an FFT-based Gammatone Frequency Cepstral Coefficient (GFCC). We use the Gaussian Mixture Model with the Universal Background Model (GMM-UBM) and a Extreme Learning Machine (ELM) as classifiers to focus on the influence of the front-ends on performance. We train the GMM-UBM and the neural network with noisy data under various conditions to investigate the impact of noise on the classifier. Our results suggest that noisy training data make an SID system noise-robust while the performance under clean conditions remains almost the same. More interestingly, training with speech-shaped noise (cocktail party) enhances SID accuracy more than white noise