Bulletin of Electrical Engineering and Informatics
Not a member yet
2885 research outputs found
Sort by
Feature selection for support vector machines in imbalanced data
Get PDFAddressing the effects of class imbalance on feature selection models has become an increasingly important focus in academic research. This study introduces a novel support vector machine (SVM)-based algorithm specifically designed to handle class imbalance during the feature selection process. Using the Taiwan bankruptcy dataset as a case study, the algorithm incorporates the ExtraTreeClassifier() to manage class imbalance and identify a reduced set of relevant variables. To validate the selected features, SVM is applied within the imbalanced data context. Subsequently, analysis of variance (ANOVA) ranking is employed to further refine the variable set to three key features. An SVM model tailored for class imbalance is then constructed to assess the effectiveness of the final feature set. The proposed model significantly outperforms existing approaches in terms of classification performance. Specifically, it achieves a Type I error of 1.17% and a Type II error of 22.9%, compared to 4.4% and 39.4% reported in prior research. In terms of overall accuracy, our method reaches 83.1%, surpassing the 81.3% achieved by earlier studies. These results demonstrate that the proposed feature selection algorithm not only improves SVM accuracy but also outperforms other feature selection techniques when used in conjunction with SVMs, particularly under conditions of class imbalance
Optimal power control for wind/solar hybrid energy system based on multi-objective particle swarm optimization
Get PDFThe effectiveness of wind and solar energy as electricity generators is significantly impacted by unpredictable and varied environmental circumstances, which affect the output power of the wind-solar hybrid power generation system. So, a control system is required for the optimal power production of hybrid renewable energy systems (HRES). This study delineates optimal power management in wind/solar hybrid energy systems by the application of multi-objective particle swarm optimization (MOPSO) algorithms, inverter controllers, and battery controllers. The MOPSO algorithm enhances power generation by modifying the duty cycle of the direct current (DC)/DC converter based on the output from the wind turbine and photovoltaic (PV) system. The proportional-integral (PI) controller functions as both an inverter and battery controller to ensure the constancy of the DC link voltage and output power. The efficacy of the developed control was evaluated using simulation. A comparison has been conducted between the efficacy of the MOPSO algorithm and the perturb and observe (PO) approach. The simulation findings indicate that the MOPSO algorithm surpasses the PO method for performance and output power. The output power produced by HRES with the MOPSO algorithm exceeds that of the PO approach. Optimal power control utilizing MOPSO can yield optimal power despite fluctuations in wind and solar intensity
Adaptive micro strip antennas for 5G networks
Get PDFThe advent of fifth-generation (5G) technology and progressing further to six-generation (6G) technology has created a new era of high-speed wireless communication, demanding antennas with enhanced capabilities to fulfill the dynamic demands of various applications. This paper presents novel approaches to designing antennas in the GHz frequency range for 5G networks by incorporating re-configurability features. Adaptive antennas provide the flexibility to alter their radiation configurations, frequencies, or polarization states, allowing them to optimize performance under different operating conditions. The theoretical foundations are explored, and reconfigurable antennas are simulated using HFSS, focusing on frequency and pattern variation at GHz frequencies using different types of switches such as pin diodes and rods. Through simulations, the antenna's S parameters are evaluated, demonstrating its capacity to meet the rigorous specifications of 5G applications. Its adaptive nature enhances connectivity and overall network performance, supporting the successful deployment and advancement of 5G technology in diverse real-world applications
BiLSTM OptiFlow: an enhanced LSTM model for cooperative financial health forecasting
Get PDFThis paper presents bidirectional long short-term memory (BiLSTM) OptiFlow, an optimized deep learning model designed to predict the financial health of cooperatives using key financial ratios: debt to equity ratio (DER), net profit margin (NPM), and return on equity (ROE). By leveraging a BiLSTM architecture combined with an optimal decayed learning rate, this model aims to enhance forecasting accuracy. The proposed model was tested against three established methods—recurrent neural network (RNN), long short-term memory (LSTM), and gated recurrent unit (GRU)—and evaluated using mean absolute error (MAE), mean absolute percentage error (MAPE), and mean squared error (MSE) metrics. Results indicate that BiLSTM OptiFlow outperforms the other models across all key indicators. This research offers a robust approach to cooperative financial forecasting, with significant implications for decision-making processes in cooperative management
Improved imperceptible engagement-based 2D sigmoid logistic maps, Hill cipher, and Kronecker XOR product
Get PDFImage encryption is a crucial facet of secure data transmission and storage, and this study explores the efficacy of combining sigmoid logistic maps (SLM), Hill cipher, and Kronecker's product method in enhancing image encryption processes. The evaluation, conducted on diverse images such as Lena, Rice, Peppers, Cameraman, and Baboon, unveils noteworthy findings. The Lena image emerges as the most successfully encrypted, as evidenced by the lowest mean squared error (MSE) at 92.81 and the highest peak signal-to-noise ratio (PSNR) at 19.43, reflecting superior fidelity and quality preservation. Additionally, the encryption of 64×64 pixels images consistently demonstrate robustness, with a high number of pixels change rate (NPCR) and unified average change intensity (UACI) values, particularly notable for the Cameraman image. Even for 128×128 pixels images, commendable encryption performance persists across the tested images. The amalgamation of SLM, Hill cipher, and Kronecker's product emerges as an effective strategy for balancing security and perceptual quality in image encryption, with the Lena image consistently outperforming others based on comprehensive metrics. This research provides valuable insights for future studies in the dynamic domain of image encryption, emphasizing the potential of advanced cryptographic techniques in ensuring secure multimedia communication
An economical approach of structural strength monitoring utilizing internet of things
Get PDFIn the current environment, structural health monitoring (SHM), has become increasingly important. The cost of sensors and connectivity has significantly decreased, allowing for remote data gathering for critical analysis and structure monitoring. This allows for the assessment and improvement of the structures' residual lifespan. The internet of things (IoT) is a network of intelligent sensors that combines the identification and detection followed by sending the different structural responses to remote computers for further analysis i.e., processing and monitoring. In this work, an integrated IoT platform for damage detection is proposed which includes an Arduino, Wi-Fi module, and sensors. The sensors gather responses from the host structure which follows a precise mathematical model is introduced to determine and measure the structural damage in comparison to the reactions of the structural member that is in good health. To determine the degree of damage, the responses recorded from the damaged and healthy beams are analyzed using the cross-correlation (CC) damage index. Moreover, the analysis carried out reveals the CC values are uploaded to the cloud, where, if the CC value is over the threshold limit, a mobile warning message is delivered
Structure of 6-dimensional finite non-commutative algebras with many single-sided units
Get PDFFinite Associative Noncommutative Algebras (FANAs) have gained considerable attention as a key foundational element for post-quantum (PQ) public-key (PK) cryptosystems, particularly those with a hidden group. These systems exploit the complexity of the hidden discrete logarithm problem (HDLP) and the challenge of solving large system of power equations. The structure of 6-dimensional FANAs over the finite field GF(p), which can include global single-sided units in different configurations (p2, p3, and p4), plays an essential role in assessing the security of these cryptosystems. A novel PQ signature algorithm has been proposed based on FANAs with p2 global single-sided units, while the others have been deemed less suitable for supporting the proposed algorithm. The decomposition of these algebras into isomorphic subalgebras, each with a global two-sided unit, significantly contributes to understanding the design of PQ cryptosystems that use FANAs with a large number of global singlesided units as their algebraic framework.Â
No binding machine learning architecture for SDN controllers
Get PDFAlthough software-defined networking (SDN) has improved the network management process, but challenges persist in achieving efficient load balancing among distributed controllers. Present architectures often suffer from uneven load distribution, leading to significant performance deterioration. While dynamic binding mechanisms have been explored to address this issue, these mechanisms are complex and introduce a significant latency. This paper proposes SDNCTRLML , a novel approach that applies machine learning mechanisms to improve load balancing. SDNCTRLML introduces a scheduling layer that dynamically assigns flow requests to controllers using machine learning scheduling algorithms. Unlike previous approaches, SDNCTRLML integrates with the standard SDN switches and adapts to different scheduling algorithms, minimizing disruption and network delays. Experimental results show that SDNCTRLML has outperformed static-binding controllers models without adding complexities of dynamic-binding systems
Tri-level lung cancer classification via deep learning based GoogleNet with computed tomography images
Get PDFLung cancer (LC) is one of the most prevalent causes of cancer-related death worldwide. World Health Organization (WHO) classifies LC into two broad histological subtypes: non-small cell lung cancer (NSCLC) which is the cause of about 85% of cases and small cell lung cancer (SCLC) which makes up the remaining 15%. Several issues can influence LC detection including poor image quality, insufficient training data, low-quality image characteristics, and poor tumor localization. To overcome these challenges a novel TRI-level LC classification via deep learning-based GoogleNet with computed tomography (CT) images (TRI-LCNet) approach has been proposed for early-stage LC detection using CT images. Initially, the LC-input images CT are collected from openly accessible datasets. The lung CT images have been preprocessed using a Gaussian star filter (GaSF) to decrease noise, followed by feature extraction using GoogleNet. The extracted LC features are then given into a support vector machine (SVM) which is utilized as a classification tool to distinguish between different classes of LC cases. The TRI-LCNet approach performance was assessed by several metrics: specificity, accuracy, F1 score, and recall. The outcomes show that the suggested method obtains a higher accuracy range of 96.93% for the early identification of LC
Energy and path loss analysis of wireless sensor networks on a robotic body (WSRobotic)
Get PDFThe objective of this work is to simulate and mathematically model both path loss and transmitted energy in a robotic wireless sensor network (WSN). The simulation and analysis showed an increase in both path loss and transmitted energy as a function of distance. The correlation between transmitted energy and path loss proved to be exponential relationship with both logarithmic and power relationships between path loss and distance. Both expressions describing path loss, using close-in (CI) dual model and transmitted energy, using wireless body area network (WBAN) model, are modified and combined in one single expression to enable optimization of energy management. The newly developed expression is simulated and produced reliable results, relating effect of frequency and message size on transmitted energy as a function of distance. Combining these results with the results showing effect on path loss on transmitted energy, enables a better optimization of energy management of nodes on robotic body. The main objective of this work, which is the development of a single expression relating transmitted energy to critical parameters (frequency, path loss exponent, message size, distance) is achieved and is logically derived and based on analysis using two separate models for path loss and transmitted energy