Indonesian Journal of Electrical Engineering and Computer Science
Not a member yet
9109 research outputs found
Sort by
Comparative analysis of machine and deep learning algorithms for semantic analysis in Iraqi dialect
Get PDFText analytics, an essential component of artificial intelligence (AI) applications, plays a pivotal role in analyzing qualitative sentiments and responses in questionnaires, particularly for governmental and private organizations. Utilizing sentiment analysis enables a comprehensive understanding of people’s opinions, especially when expressed in lengthy texts in their native language, with minimal constraints. This study aims to identify the determinants of electronic service adoption among Iraqi citizens. A set of 1,695 questionnaires were distributed to Iraqi citizens; obtained 1,234 responses that were increased via data augmentation to 1,393 comments. Four machine learning (ML) and three deep learning (DL) algorithms Na¨ıve Bayes (NB), K-nearest neighboror machine (SVM), random forest (RF), as well as two variants of long-shortterm memory (LSTM) networks and convolutional neural networks (CNN) were employed to classify qualitative feedback. Following rigorous training and testing, the NB classification algorithm exhibited the highest accuracy, achieving 82.89%
Adversarially robust federated deep learning models for intrusion detection in IoT
Get PDFEnsuring the robustness, security, and privacy of machine learning is a pivotal objective, crucial for unlocking the complete potential of the internet of things (IoT). Deep neural networks have proven to be vulnerable to adversarial perturbations imperceptible to humans. These perturbations can give rise to adversarial attacks, leading to erroneous predictions by deep neural networks, particularly in intrusion detection within the IoT environment. This paper introduces a federated adversarial learning framework designed to protect both data privacy and deep neural network models. This framework consists of federated learning for data privacy and adversarial training on IoT devices to enhance model robustness. The experiments show that adversarial training at the Fog node devices significantly improves the robustness of a federated learning model against adversarial attacks when compared to normal training. Furthermore, the proposed adversarial deep federated learning model is validated using the Edge-IIoTset dataset, achieving an accuracy rate of 91.23% in the detection of attacks
Machine learning based prediction of production using real time data of a point bottom sealing and cutting machine
Get PDFThe packaging sector utilizes polypropylene based flexible materials for diverse product packaging with customization options in size and design achieved through advanced flexographic printing and point bottom sealing and cutting machines. Accurately estimating production time and quantity is vital for efficient planning and cost estimation, with factors like material dimensions, thickness, and cutting machine speed influencing production output. Understanding the intricate relationship between these parameters is essential for comprehending their impact on production time and quantity. Predicting production quantity before production begins helps in determining machine runtime and associated costs. In large-scale production systems, machine learning (ML) has proven to be a useful tool for resource allocation and predictive scheduling. An attempt has been made in this paper to develop an intelligent model for predicting the yield of a cutting machine using artificial neural network (ANN), support vector regression (SVR), regression tree ensemble (RTE) and gaussian process regression (GPR). The most crucial features for prediction were identified and the hyperparameters of the ML models were optimized to create efficient models for prediction. A comparative analysis of the four models revealed that the GPR model was simple and effective with least training time and prediction error
Designing stair climbing wheelchairs with surface prediction using theoretical analysis and machine learning
Get PDFUrban settings present considerable obstacles for those use personal mobility wheelchairs, especially when it comes to manoeuvring stairs. The objective of this study is to improve the safety and ease of use of wheelchairs designed for ascending stairs. The study aims to tackle the significant issue of instability and limited ability to adjust to different types of terrain. This research employs a holistic methodology that combines theoretical dynamic analysis, hardware design and simulation, and field testing, in addition to advanced machine learning approaches for surface prediction. Theoretical models guarantee the stability of the wheelchair, while hardware simulations offer valuable insights into its structural integrity. The data obtained from inertial measurement unit (IMU) sensors during field tests is analysed and categorised using models like random forest and gradient boosting, which exhibit exceptional accuracy in forecasting movement circumstances. The results demonstrate that the implementation of these combined techniques greatly enhances the wheelchair’s capacity to safely manoeuvre over urban barriers. The study finds that the suggested solutions show great potential for creating intelligent mobility aids, which might be used to improve accessibility for those with mobility limitations
Medical image registration and classification using smell agent rat swarm optimization based deep Maxout network
Get PDFMedical image registration (MIR) is a crucial task in clinical image processing, involving the alignment of images from different modalities, such as magnetic resonance imaging (MRI) and computed tomography (CT), across various time points and subjects. Despite numerous advancements, no universal method caters to all MIR applications. This paper introduces the smell agent rat swarm optimization based deep Maxout network (SARSO-DMN) for MIR and classification. This work aims to enhance the accuracy and efficiency of medical image alignment and classification, addressing the challenges posed by diverse imaging modalities and temporal variations. The problem involves effectively registering CT and MRI images, followed by inhale and exhale classification. The proposed approach begins with feeding the input images into a convolutional neural network (CNN), followed by applying a deformation field to generate an intermediate output (output-1). This output, along with the input MRI images, is further processed by a CNN to produce output-2. Subsequently, output-2 and the input MRI image are subjected to another CNN, resulting in the final registered image. The classification phase utilizes a DMN optimized by the SARSO algorithm, which combines smell agent optimization (SAO) and rat swarm optimizer (RSO). The results demonstrate that SARSO-DMN achieves a maximum accuracy of 90.7%, a minimum false positive rate (FPR) of 11.3%, and a maximum true positive rate (TPR) of 91.2%. The SARSO-DMN approach provides a robust solution for MIR and classification, leveraging advanced optimization techniques to enhance performance
G2M weighting: a new approach based on multi-objective assessment data (case study of MOORA method in determining supplier performance evaluation)
Get PDFCriteria weighting methods in decision support system (DSS) face various challenges and limitations that can affect their accuracy and reliability. One of the main challenges is subjectivity, this subjective assessment can reduce the objectivity and consistency of results. The main objective of the new weighting method grey geometric mean (G2M) weighting is to provide more objective and robust criteria weights under conditions of uncertainty and incomplete data. The new G2M weighting approach has a significant potential impact on the DSS field, it has the potential to generate more effective and efficient decisions, which can improve organizational performance, reduce risk and optimize outcomes. Pearson correlation test results of two sets of rankings generated by DSS methods namely grey relational analysis (GRA), simple additive weighting (SAW), multi-attributive ideal-real comparative analysis (MAIRCA), weighted product (WP), combined compromise solution (COCOSO), vlsekriterijumska optimizacija i kompromisno resenje (VIKOR), and a new additive ratio assessment (ARAS) that there is a strong positive correlation between the two methods using G2M weighting criteria. The high correlation value indicates that the rankings of the methods used tend to move together, giving confidence in the consistency and validity of the resulting ranking results. This gives confidence that both methods can be used simultaneously or interchangeably with consistent results. The use of G2M weighting in the DSS method used can support better decision-making by providing consistent information and validity of ranking results
Comparative study of pothole detection using deep learning on smartphone
Get PDFPotholes present a significant problem in many countries, leading to vehicle damage and traffic accidents. These road imperfections pose safety risks and impose economic burdens. Despite existing detection methods using sensors and computer vision deep learning processed on PCs, a gap remains in deploying cost-effective, widely accessible solutions. This study aims to bridge this gap by developing deep learning models optimized for smartphones, reducing costs and enhancing deployment feasibility. We developed multiple models for pothole detection, utilizing transfer learning and Bayesian hyperparameter tuning to optimize detection accuracy and resource efficiency. Our evaluations focused on computationally light models such as YOLOv8 small, YOLOv8-nano, YOLOv7 tiny, and faster R-CNN MobileNetV3. In terms of detection accuracy, YOLOv8 small and YOLOv8 nano stood out, achieving average precisions (AP) of 83.5% and 82.5%, respectively. YOLOv8 nano proved the most efficient, offering high detection accuracy, a file size three times smaller than YOLOv8 small in TFLite format, and the fastest inference time of 0.72 seconds per image. This study highlights the potential of smartphones in urban pothole detection, contributing to improved road maintenance and urban policy
An evaluation model of website testing framework based on ISO 25010 performance efficiency
Get PDFTesting is an important aspect of software development. Automation testing is now widely used to achieve better and more efficient results. Various automation testing frameworks are available in the market. However, one of the major challenges is determining which automation testing framework is suitable for testing. This study proposes an evaluation model for evaluating web automation testing frameworks based on seven performance efficiency factors to address this issue. The model evaluates five types of transactions commonly used on the web; CRUD, Get Massive Data, search, file upload, and file download. In addition, the tested frameworks are categorized as good, medium, and low. To measure the success of the research, expert weighting was also used. Based on the results obtained for all types of transactions, almost all classifications between the experimental results and weighting were in the same class. Although the model was found to be effective with a 100% accuracy rate, it had an accuracy rate of 80% for upload transactions. The outcomes of this study serve as a valuable reference for choosing suitable software for both tested frameworks and other software applications. In future studies focus on narrowing the selection based on not only performance but also functionality and ease of use
Identification of chilli leaf disease using contrast limited histogram equalisation and k-means clustering
Get PDFPlant disease diagnosis is crucial for preventing productivity and quality losses in agricultural products. Because plants are continually attacked by insects, bacterial infections, and smaller scale organisms it is necessary for early diagnosis disease control is a vital part of profitable chilli crop production, hence early diagnosis of disease identification is an important aspect of crop management. This paper discusses strategies for detecting disease effectively in order to improve chilli plant product quality. An image processing technique based on identification of chilli leaf disease using contrast limited histogram equalisation and k-means clustering (KMC). The approach was carried out in five stages: acquiring the image, preprocessing, extracting features, classifying the diseases, and showing the outcome. This work offers a thorough implementation of CLAHE for preprocessing, k-means cluster for feature extraction and support vector machine (SVM) for classification of chilli leaf diseases. The accuracy was tested for standard chilli dataset for major 2 types of diseases including anthracnose and bacterial blight form kaggle dataset with varying samples of 70:30 and 60:40 respectively and it is observed that the average accuracy improved to 98% compared to existing techniques
High-accuracy classification of banana varieties using ResNet-50 and DenseNet-121 architectures
Get PDFBananas are a popular fruit in Indonesia due to their affordability, availability, and rich nutritional content. Identifying different banana types is crucial for consumption and processing, yet some types are difficult to distinguish visually. This study aims to classify banana types using convolutional neural network (CNN) architectures, specifically ResNet-50 and DenseNet-121. The dataset consists of five banana classes, which were processed using preprocessing techniques to enhance image quality prior to model training. The results demonstrate that the proposed models can classify banana types with high accuracy. The research methodology includes data collection, preprocessing, CNN model implementation, and performance evaluation using a confusion matrix. The dataset was split into training and testing sets in an 80:20 ratio, with validation data extracted from the training set in a 90:10 ratio. The models were trained on the training data, validated with validation data, and tested on the testing data to assess final performance. The study concludes that the CNN architectures employed are effective in classifying banana types, with the DenseNet-121 model achieving 93.02% accuracy, outperforming the ResNet-50 model, which achieved 92.44%. These results indicate that the models can capture essential features from banana images and produce accurate predictions