Indonesian Journal of Electrical Engineering and Computer Science
Not a member yet
9109 research outputs found
Sort by
Privacy preserving ZK-STARK based blockchain for agriculture food supply chain
Get PDFThe blockchain-based applications such as food supply chain (FSC), healthcare are becoming increasingly popular due to their decentralized nature, distributed structure, and the ability to track products. Still, there are concerns regarding the privacy of transacted data, including personal identities, as all transactions are recorded on a ledger accessible to participating nodes. Existing technologies of privacy preservation are vulnerable to quantum attacks, which will pose a significant threat to blockchain applications in the future. To address this, a proposed model uses modified zero-knowledge scalable transparent argument of knowledge (ZK-STARK) in the blockchain FSC by utilizing three different polynomial interpolation methods. Performance measurements indicate that the fast fourier transform (FFT) outperforms the others. Unlike ZK-SNARK, ZK-STARK does not require a trusted setup, makes scalable and transparent. By defending against quantum attacks, this model enhances the security of the blockchain system. The blockchain-based FSC is implemented using hyperledger composer, with all entities completing transactions privately through ZK-STARK and smart contracts. But, ZK-STARK may add performance overhead into the blockchain FSC. Future work will aim to reduce the performance overhead of ZK-STARK, decide which operations should be off-chain or on-chain, and compare the performance of this new model to the existing system
Multi-class chronic lung disease classification based on guided backpropagation convolutional neural network using chest X-ray images
Get PDFClinical diagnosis is crucial as chronic lung disease is a leading cause of mortality worldwide. Chest X-ray imaging is essential for the early and accurate diagnosis of lung diseases. However, due to the complexity of pathological abnormalities and detailed annotation, the computer-aided diagnosis of lung diseases is challenging. To overcome this challenge, this research proposes the guided backpropagation convolutional neural network (GBPCNN) for the classification of chronic lung disease into 14 classes, by adjusting the network’s weights in CNN layers. The GBP technique enhances result accuracy by pinpointing the regions in an input image. Initially, the chest X-ray radiography (CXR) dataset is collected for estimating the effectiveness of the classifier. After collecting the dataset, the pre-processing is performed by utilizing image denoising using gaussian filter and normalization techniques. Then, the pre-processed data is fed to the feature extraction process, and it is done by using EfficientNetB2. Finally, extracted features are provided to the classification process to categorize chronic lung disease into 14 classes. The experimental results show that the proposed GBPCNN method attains better results and it achieves the accuracy of 97.94% as compared to the existing approaches like MobileLungNetV2 and CX-Ultranet. These results highlight the potential of our approach for clinical applications
Machine learning-based classification of corn seed viability using electrical impedance spectroscopy
Get PDFCorn (Zea mays L.), an essential global commodity, plays an ever-increasing role in agri-food systems. To support growing demand, rapid and noninvasive methods for determining seed germination rates are crucial alongside invasive techniques such as dissection, germination paper tests, and chemical assays. This study introduces electrical impedance spectroscopy (EIS) as a novel, non-invasive approach for classifying viable and non-viable corn seeds. Non-viable corn seeds were prepared by exposing them to 100 °C convection heat for 30 minutes. Impedance spectra were measured using the EVAL-AD5933EBZ evaluation board from 400 kHz to 1 MHz frequency range within 30 seconds. Furthermore, a comparison of six optimized supervised machine learning (ML) algorithms, including shallow and deep networks, was performed, setting this apart from other studies. The trained model was deployed to assess the viability of new seed samples effectively. Key impedance metrics, including their frequencies, were extracted to train and test the algorithms. The regression tree (RTree) model outperformed deep learning classifiers, achieving 95% accuracy, 90% precision, and 100% sensitivity. The results indicated an upward trend in viable seed impedance, increasing by 0.000164 Ω/Hz, peaking at 990 kHz. This approach offers a rapid, non-invasive solution for seed viability assessment, with significant potential to enhance agricultural productivity
Detection of diabetic retinopathy and classification of its stages by using convolutional neural network
Get PDFDiabetes detection is pivotal in disease management and complication prevention. Traditional screening methods, like blood tests, are invasive and time-consuming. Deep learning has emerged as a non-invasive and automated alternative for diabetes detection. Convolutional neural networks (CNNs) excel in image analysis tasks, making them ideal for this purpose. This paper employs a CNN-based method for diabetes prediction using retinal images, utilizing the DenseNet169 architecture for feature extraction and diabetic retinopathy (DR) prediction. The APTOS 2019 blindness detection dataset from Kaggle, containing around 13,000 retinal images, is used for training. Pre-processing and normalization precede feature extraction, followed by the prediction of the DR stage. The model aims to classify retinal images into five stages of DR (0 to 4), ranging from no DR to proliferative DR. Our model achieved over 82% accuracy, outperforming advanced algorithms. Model evaluation includes accuracy, precision, recall, and F1 score measures
Machine learning based stator-winding fault severity detection in induction motors
Get PDFApproximately 35% of all induction motor defects are caused by stator inter-turn faults. In this paper a novel algorithm has been proposed to analyze the three-phase stator current signals captured from the motor while it is in operation. The suggested method seeks to identify stator inter-turn short circuit faults in early stage and take the appropriate action to prevent the motor's condition from getting worse. Three-phase current signals have been captured under healthy and faulty conditions of the motor. Involving discrete wavelet transform (DWT) based decomposition followed by reconstruction using inverse DWT (IDWT), 50 Hz fundamental component has been removed from the captured raw current signals. Subsequently, from each phase current 15 statistical parameters have been retrieved. The statistical parameters include mean, standard deviation, skewness, kurtosis, peak-to-peak, root mean square (RMS), energy, crest factor, form factor, impulse factor, and margin factor. At the end, a standard machine learning algorithm namely error correcting output codes-support vector machine (ECOC-SVM) has been employed to classify six different severity of stator winding faults. The proposed fault diagnosis method is load and motor-rating independent
Natural smart home automation system using LSTM based on household behaviour
Get PDFA smart home automation system (SHAS) utilizing data-driven learning is an advanced internet of things (IoT) application aimed to learn household behavior to prevent miniatur circuit breaker (MCB) trips due to overload. Unlike traditional deterministic methods, this study leverages a layered AI model, featuring real-time data collection, long short-term memory (LSTM) based learning, and an automatic control system. The LSTM classification model generates precise ON/OFF control signals sent to IoT smartplugs, optimizing appliance usage and reducing the risk of electrical overload. Data from smartplug sensors, including appliance status and environmental factors like power consumption, temperature, and humidity, were collected every minute over three months, yielding 80,818 data points. The system's performance was evaluated on three appliances: Air Conditioner, Television, and Water Pump Machine. Results showed high accuracy for Television at 98% and Water Pump Machine at 97.6%, with slightly lower accuracy for Air Conditioner at 81.9%. This demonstrates the system's effectiveness in real-world applications. The scalability and adaptability of the Natural SHAS model to different appliances and environments mark a significant advancement in smart home automation, offering a practical solution for preventing electrical overload and improving household energy management
Optimizing FBMC/OQAM: Hermite filter and DFT-based precoding for PAPR reduction
Get PDFIn the ever-evolving landscape of wireless communication, there is a persistent quest for modulation schemes that optimize spectral efficiency, reduce interference, and enhance overall system performance. This paper introduces a novel modulation technique that synergistically improves on the strengths of filter bank multi-carrier (FBMC). A distinctive feature of our approach is the deployment of the Hermite prototype filter in the FBMC system, diverging from traditional FBMC architectures. An advanced precoding strategy leveraging a pruned discrete fourier transform (pDFT) paired with scaling is also introduced. This combination promises reduced inter-symbol interference and heightened spectral efficiency. As the management of the peak-to-average power ratio (PAPR) is a significant challenge in FBMC systems to addressing this iterative particle swarm optimization (IPSO) algorithm is proposed. Evaluations are carried out to demonstrate the efficiency of the proposes scheme in reducing PAPR substantially for FBMC/OQAM framework. Experiments are conducted and comparisons are performed among several prominent multicarrier modulation schemes. The results from the experiments indicate that the application of IPSO algorithm with Hermite functions and applied to an FBMC/OQAM system using pruned DFT has been successful in reducing the PAPR also a 6-13% decrease in error rate has been shown across varying QAM orders regardless of SNR level
Detecting network security incidents in wireless sensor networks using machine learning
Get PDFThis study enhances the domain of cybersecurity within wireless sensor networks (WSNs) through the integration of sophisticated artificial intelligence (AI) and machine learning (ML) techniques. By conducting an exploratory data analysis (EDA), this research reveals critical insights into network behavior, facilitating the development of predictive models for anomaly detection. The application of ML algorithms decision trees (DT) and random forest (RF) demonstrated dominant performance in identifying potential security threats, as evidenced by metrics accuracy, precision, recall, and F1 scores. This work not only enhances the security framework for WSNs but also contributes to the extensive field of network security, offering a robust analytical and predictive methodology for future cybersecurity initiatives. The advanced model can be deployed in other WSN and internet of things (IoT) based applications
Predicting peak demand for electricity consumption using time series data and machine learning model
Get PDFEnergy consumption is influenced by various factors, including the proliferation of electronic devices, technological advancements, economic growth, agricultural development, and population increase. Each of these factors contributes to the rising demand for energy. This paper addresses the challenge of predicting peak energy demand (ED) by utilizing historical time series data from the past five years, combined with temperature data from Tamil Nadu’s official sources. We employed feature engineering techniques to prepare the data for machine learning models, specifically XGBoost regressor, lasso, and ridge regression. The time series data was then analyzed using both univariate and multivariate models, including auto regressive integrated moving average (ARIMA) and vector autoregressive (VAR) models. The results show that our models can effectively forecast ED, providing critical insights for policymakers and stakeholders involved in energy planning and resource management
Efficient deep learning approach for enhancing plant leaf disease classification
Get PDFThe widespread occurrence of plant diseases is a major factor in the reduction of agricultural output, affecting both crop quality and quantity. These diseases typically begin on the leaves, influenced by alterations in plant structure and growing techniques, and can eventually spread over the entire plant. This results in a notable decrease in crop variety and yield. Successfully managing these diseases depends on accurately classifying and detecting leaf infections early, which is essential for controlling their spread and ensuring healthy plant growth. To address these challenges, this paper introduces an efficient approach for detecting plant leaf diseases. A concatenation of pre-trained convolutional neural networks (CNN) for enhanced plant leaf disease using transfer learning technique is implemented, with a specific focus on accurate early detection, utilizing the comprehensive new plant diseases dataset. The combined residual network-50 (ResNet-50) with densely connected convolutional network-121 (DenseNet-121) architecture aims to provide an efficient and reliable solution to these critical agricultural concerns. Various evaluation metrics were utilized to evaluate the robustness of the proposed hybrid model. The proposed ResNet-50 with the DenseNet-121 hybrid model achieved a rate of accuracy of 99.66%