Indonesian Journal of Electrical Engineering and Computer Science
Not a member yet
9109 research outputs found
Sort by
Deep learning-based stacking ensemble for malaria parasite classification in blood smear images
Get PDFMalaria remains a significant global health challenge, necessitating accurate and efficient diagnostic tools. Deep learning models have emerged as promising solutions for automated malaria detection using microscopic blood smear images. This study evaluates the performance of various convolutional neural network (CNN) architectures, including VGG16, ResNet50, MobileNetV2, and EfficientNet, in classifying infected and uninfected cells. Individual model performances were assessed based on accuracy, precision, recall, and F1-score, with EfficientNet achieving the highest standalone accuracy of 88.0%. To enhance classification performance, a stacking ensemble approach was implemented, using a logistic regression meta-classifier to integrate outputs from multiple models for improved decision-making. The stacking model outperformed individual networks, achieving an accuracy of 89.4%, with precision, recall, and F1- scores surpassing those of standalone models. Challenges in malaria parasite classification—such as high inter-class similarity, variations in staining quality, and class imbalance were addressed through data augmentation and model tuning. These findings highlight the potential of ensemble learning in medical image analysis, paving the way for more accurate and scalable malaria detection systems
Comparative analysis of Cohen-Coon and Ziegler-Nichols tuning methods for three-phase induction motor with speed sensorless control
Get PDFThe use of speed sensors in the speed controller of three-phase induction motors affects the reliability of the induction motors. In addition, the drive engine that is often used in industry is a three-phase induction motor. So, speed sensorless control is needed for induction motors to achieve the best performance. This study uses a discrete disturbance observer (D0) as feedback on the speed sensorless control. The controller used in this method is a discrete PI with the Cohen-Coon (CC) and Ziegler-Nichols (ZN) tuning method. The purpose of this study is to obtain a comparative analysis of the CC and ziegler nichols tuning method using a discrete PI on the speed sensorless control scheme with torque load variation. This study was carried out experimentally using an Alliance AY3A-90L4 induction motor. The results show that the CC tuning method is better under parameter efficiency and robustness against disturbance and ZN is better under parameter reliability
Performance evaluation of transdermal optical wireless communication using spatial diversity techniques
Get PDFActive medical implants and other contemporary medical applications need a dependable, high-speed communication channel between external and internal transceivers. Optical wireless communications have demonstrated advantages over widely used radio frequency technology in terms of data speeds, bandwidth abundance, and immunity to interference. Regretfully, this advantage implies strict alignment requirements for both sending and receiving ends. This study focuses on the effects of using multiple transmitters or receivers under the influence of pointing error on the transcutaneous link's overall performance measured by the outage probability and outage rate. Spatial diversity techniques have demonstrated their viability in increasing the link's reliability in free space optical communications. This drives the investigation of improvement transdermal communication system by adding numerous transmitters or receivers. Various misalignment severities are used to represent different operating circumstances, and these analyses result in explicit closed-form formulas for the relevant metrics. The findings clearly show the benefits of employing multiple transmitters and receivers on the link's outage performances. A notable improvement in the average signal-to-noise ratio values for the outage probability and outage rate compared to the single input single output setup was achieved. Furthermore, the theoretical conclusions are subsequently confirmed by MATLAB-based Monte-Carlo simulation for several instructive cases
Optimizing energy efficiency and improved security in wireless sensor networks using energy-centric MJSO and MACO for clustering and routing
Get PDFWireless sensor networks (WSNs) play a pivotal role in various applications, but their energy-constrained nature poses significant challenges to their sustainable operation. In this paper, we propose a novel approach to enhance energy efficiency in WSNs by leveraging energy-centric multi-objective jaya search optimization (MJSO) and multi-objective ant colony optimization (MACO) for clustering and routing. Our method aims to address the energy consumption issues by optimizing clustering and routing strategies simultaneously. The energy-centric MJSO algorithm is employed to intelligently organize sensor nodes into clusters, considering energy consumption, network coverage, and connectivity. The multi-objective MACO algorithm optimizes routing paths by balancing energy consumption and network lifetime objectives. Through integration and simulations, the approach enhances energy efficiency in WSNs for various applications like environmental monitoring and smart cities, advancing energy-efficient clustering and routing. By integrating energy-centric MJSO and MACO into clustering and routing protocols, WSNs can achieve significant improvements in energy efficiency and security while maintaining reliable communication and data delivery
An approach-based ensemble methods to predict school performance for Moroccan students
Get PDFEducation is a key factor in Morocco's development, with school performance serving as a critical measure of the education system’s quality. However, disparities in student outcomes remain, influenced by socioeconomic, demographic, and infrastructural factors. Our study aims to develop a predictive model to assess and improve school performance in Morocco using ensemble machine learning techniques, focusing on the stacking approach. Data from the Massar platform includes variables such as gender, age, type of school, parental occupation, academic results, and residential area. After rigorous data cleaning and preprocessing, a stacking model was created by combining predictions from five base models: random forest, gradient boosting, k-nearest neighbors (KNN), support vector machine (SVM), and multi-layer perceptron (MLP). A random forest metamodel was used to integrate these results. The experimental results of the paper demonstrate the effectiveness of our approach. The stacking model achieved an accuracy of 78.70%, surpassing the individual base models. The meta-model demonstrated strong reliability, achieving an F1 score of 78.62% while reducing false negatives and ensuring balanced predictions. Among the base models, neural networks showed the best performance, achieving the highest predictive accuracy. This research highlights the potential of stacking methods for predicting school performance. Incorporating additional variables, such as parental education and teacher attributes, could further refine the model and enhance Morocco’s educational outcomes
Wolfram Alpha based-inventory model for damaged items of pharmaceutics by utilizing exponential demand rate
Get PDFIn this study, an inventory model is developed for pharmaceutical products that deteriorate over time with an exponential demand rate. The discussion of exponential demand is rarely explored but has the advantage that the demand value toward total cost remains positive. This study assumes allowable shortages and complete backlogging, making it necessary to design an optimal policy for deteriorating goods with an exponential demand rate. The model shows that the initial stock decreases over time, potentially leading to shortages before the next order arrives. The optimal solution indicates that the inventory reaches the zero point at 1 = 0.0000011 and the cycle length 1 = 0.012 resulting in an average minimum total cost of ̅̅̅̅ = $17,133.9 per cycle by Wolfram Alpha. Sensitivity analysis measures the changes of the results in the increasing value of ̅̅̅̅ for all parameters. Exponential function variables ( and ) produces 1 and 1 stable values. On increasing the cost of each damage () and constant damage rate () produces a 1 stable value, but the value of 1 increases. An increase in storage costs (h) results in a decrease in the value of 1 and 1. Increasing in the cost of shortages (s) resulted in an increase in the value of 1 and a decrease in the value of 1
Recognizing AlMuezzin and his Maqam using deep learning approach
Get PDFSpeech recognition is an important topic in deep learning, especially to Arabic language in an attempt to recognize Arabic speech, due to the difficulty of applying it because of the nature of the Arabic language, its frequent overlap, and the lack of available sources, and some other limitations related to the programming matters. This paper attempts to reduce the gap that exists between speech recognition and the Arabic language and attempts to address it through deep learning. In this paper, the focus is on Call for Prayer (Aladhan: ناذآلا ) as one of the most famous Arabic words, where its form is stable, but it differs in the notes and shape of its sound, which is known as the phonetic Maqam (Maqam: ماقملا يتوصلا ). In this paper, a solution to identify the voice of AlMuezzin ( نذؤملا ), recognize AlMuezzin, and determine the form of the Maqam through VGG-16 model presented. The VGG-16 model examined with 4 extracted features: Chroma feature, LogFbank feature, MFCC feature, and spectral centroids. The best result obtained was with chroma features, where the accuracy of Aladhan recognition reached 96%. On the other hand, the classification of Maqam with the highest accuracy reached of 95% using spectral centroids feature
Enhancing TV program success prediction using machine learning by integrating people meter audience metrics with digital engagement metrics
Get PDFWith the emergence of numerous media services on the internet, television (TV) remains a highly demanded medium in terms of reliability and innovation, despite intense competition that compels us to devise strategies for maintaining audience engagement. A key indicator of a TV channel’s success is its reach, representing the percentage of the target audience that views the broadcasts. To aid TV channel managers, the industry is exploring new methods to predict TV reach with greater accuracy. This paper investigates the potential of advanced machine learning models in predicting TV program success by integrating people meter audience metrics with digital engagement metrics. Our approach combines convolutional neural networks (CNNs) for processing digital engagement data, long short-term memory (LSTM) networks for capturing temporal dependencies, and gaussian processes (GPs) for modeling uncertainties. Our results demonstrate that the best-performing hybrid model achieves a prediction accuracy of 95%. This study contributes to the field by addressing manual scheduling errors, financial losses, and decreased viewership, providing a more comprehensive understanding of audience behavior and enhancing predictive accuracy through the integration of diverse data sources and advanced machine learning techniques
HangeulVR: an immersive and interactive Korean alphabet learning on virtual reality
Get PDFLearning a new foreign language promises numerous benefits, such are career advantage, culture exposure, and traveling opportunity. However, it comes with a cost of considerably significant efforts and time commitment. The challenge intensifies when dealing with languages characterized by distinctive scripts, such as Hangeul in Korean language. The requisite mastery of Hangeul characters precedes the exploration of fundamental linguistic elements, including grammar, pronunciation, speaking, and writing. In this research, we propose an innovative, immersive, and interactive methodology for Hangeul acquisition employing virtual reality (VR). Our study transports participants into a virtual environment, guided by a gamification framework designed to facilitate Hangeul learning. Participants are able to learn basic pronunciation, listening, and Hangeul writing, three fundamental aspects of learning the Korean alphabet. Empirical findings from our experiments show the potential of its usage, indicated by its system usability scale (SUS) of 74.4
Comprehensive multiclass debris detection for solar panel maintenance using ANN models
Get PDFSolar photovoltaic (PV) technology has emerged as a leading renewable energy solution globally. However, maintaining optimal performance remains a challenge due to the accumulation of debris, including dust, bird droppings, and other contaminants on the panels. These deposits significantly reduce the efficiency of solar panels, necessitating regular monitoring and cleaning. Automated inspection systems provide a cost-effective alternative to traditional methods by minimizing labor-intensive efforts. This study proposes a machine learning-based framework for detecting and classifying several types of debris on solar panels. The methodology utilizes gray-level co-occurrence matrix (GLCM) texture features and key statistical features extracted from RGB, HSV, and LAB color spaces. A dataset comprising 19 distinct classes, such as “Without Dust,” “Bird Droppings,” “Black Soil,” and “Sand,” was employed to train and evaluate the models. Among the tested classification techniques, artificial neural networks (ANN) achieved a notable accuracy of 93.94%, demonstrating their effectiveness in identifying and categorizing debris. This work underscores the potential of machine learning-based feature extraction and classification techniques to automate solar panel inspection and facilitate targeted cleaning interventions, thereby enhancing overall system efficiency