Indonesian Journal of Electrical Engineering and Computer Science
Not a member yet
9109 research outputs found
Sort by
A comparative study of pre-trained models for image feature extraction in weather image classification using orange data mining
Get PDFThis paper presents a detailed comparative analysis of pre-trained models for feature extraction in the domain of weather image classification. Utilizing the orange data mining toolkit, we investigated the effectiveness of six prominent pre-trained models-InceptionV3, SqueezeNet, VGG-16, VGG-19, painter, and DeepLoc-in accurately classifying weather phenomena images. Among these models, InceptionV3, in conjunction with neural networks, emerged as the most effective, achieving a classification accuracy (CA) of 96.1%. Painter and SqueezeNet also showed strong performance, with accuracies of 95.1% and 86.7%, respectively, although they were surpassed by InceptionV3. VGG-16 and VGG-19 provided moderate accuracy, while DeepLoc underperformed significantly with a maximum accuracy of 56%. Neural networks consistently outperformed other classifiers across all models. This study highlights the critical importance of selecting appropriate pre-trained models to enhance the accuracy and reliability of weather image classification systems
Design and testing of a nutrient solution control system for soilless culture using mathematical models
Get PDFThe optimization of nutrient management is crucial for successful soilless plant cultivation, where precise control of fertilizer application significantly impacts plant growth. This research addresses the challenge of developing an effective nutrient control system tailored for soilless cultivation by focusing on regulating electrical conductivity (EC) levels in nutrient solutions. The proposed system utilizes mathematical models and linear regression techniques to manage the nutrient solution mixing ratio. To ensure accuracy, sensors were calibrated, achieving a 99.59% accuracy rate for pH measurement and 95.25% for EC measurement. Experimental validation of the system demonstrated that, with a target EC range of 1.5-2.3 mS/cm, a 10 L solution volume yielded a maximum error rate of 1.75% and an average error of 0.95%. In contrast, a 50 L solution volume showed a slight increase in maximum error rate to 2.89% and an average error of 2.08%. These results highlight the system’s capability to precisely adjust EC levels using a defined linear regression model for AB liquid fertilizer ratios. In conclusion, the developed system effectively controls nutrient levels, demonstrating its potential for enhancing nutrient management in hydroponic farming applications
A novel model for detecting web defacement attacks transformer using plain text features
Get PDFOver the last decade, web defacements and other types of web attacks have been considered serious security threats to web-based services and systems of many enterprises and organizations. A website defacement attack can bring severe repercussions to the website owner, such as immediate discontinuance of the website operations and damage to the owner’s reputation, which may lead to enormous monetary losses. Several solutions and tools for monitoring and detecting web defacements have been designed and developed. Some solutions and tools are limited to static web pages, while others can handle dynamic ones but demand significant computational power. The existing proposals’ other issues are relatively low detection rates and high false alarm rates because many crucial elements of web pages, including embedded code and images are not properly processed. This paper proposes a novel model for detecting web defacements to address these issues. The model is based on the bidirectional long-short term memory (Bi-LSTM) deep learning method using features of the plain text content extracted from web pages. Comprehensive testing on over 96,000 web pages dataset demonstrates that the proposed Bi-LSTM-based web defacement detection model outperforms earlier methods, achieving a 96.04% overall accuracy and a 2.03% false positive rate
Comparative analysis of whale and Harris Hawks optimization for feature selection in intrusion detection
Get PDFThis research paper explores the efficacy of two nature-inspired optimization algorithms, the whale optimization algorithm (WOA) and Harris Hawks optimization (HHO), for feature selection in the context of intrusion detection and prevention systems (IDPS). Leveraging the NSL-KDD dataset as a benchmark, our study employs Python for implementation and uses decision tree (DT) as the classification model. The objective is to assess the impact of the HHO and WOA optimization techniques on the performance of IDPS through feature selection. The WOA and HHO techniques were able to lessen the features from 40 to 16 and 13, respectively. Results indicate that DT integrated with HHO achieves an impressive accuracy of 97.59%, outperforming the WOA-enhanced model, which attains an accuracy of 97.5%. This study contributes valuable insights into the comparative effectiveness of WOA and HHO optimization algorithms in enhancing the accuracy of IDPSs, shedding light on their potential applications in the realm of cybersecurity
Apache Spark based distributed clustering for big data analytic with application to 3D road network
Get PDFThe vast amount of data stored nowadays has turned big data analytics into a very promising research field. Clustering is an essential step in data analysis, widely used for classification, collecting statistics, and acquiring insights in specific domains of knowledge. However, the most of existing algorithms based on Lloyd-Forgy’s method, have an enormously huge average-case complexity while clustering data sets with a large number of features, which may be superpolynomial time (NP-hard) and are severely constrained in terms of speed, productivity, and adaptability. Aiming to improve Lloyd-Forgy’s clustering performance, K-means++ algorithms, a variety of algorithm-level optimizations which is not been well-studied, is discussed along with very promising gaussian mixture model (GMM) and soft clustering based Fuzzy C-means (FCM). Further, for fast and distributed data processing and to leverage the benefits of big data platforms, such as Apache Spark, Spark-based clustering methods are applied on three-dimensional (3D) road network data set which is collected from UCI repository. However, Spark-based clustering research is still in infancy. The distributed computation tests are conducted by allocating two core processors and one databricks unit (DBU) with 15 GB memory and measuring execution times, as well as root mean square error (RMSE), mean absolute error (MAE), clustering accuracy, and silhouette values. The results are promising and provide new research directions in the field of spark-based clustering on big data
A conceptual approach of optimization in federated learning
Get PDFFederated learning (FL) is an emerging approach to distributed learning from decentralized data, designed with privacy concerns in mind. FL has been successfully applied in several fields, such as the internet of things (IoT), human activity recognition (HAR), and natural language processing (NLP), showing remarkable results. However, the development of FL in real-world applications still faces several challenges. Recent optimizations of FL have been made to address these issues and enhance the FL settings. In this paper, we categorize the optimization of FL into five main challenges: Communication Efficiency, Heterogeneity, Privacy and Security, Scalability, and Convergence Rate. We provide an overview of various optimization frameworks for FL proposed in previous research, illustrated with concrete examples and applications based on these five optimization goals. Additionally, we propose two optional integrated conceptual frameworks (CFs) for optimizing FL by combining several optimization methods to achieve the best implementation of FL that addresses the five challenges
Design of stress detector with fuzzy logic method (GSR and heart rate parameters)
Get PDFStress is a condition of tension that affects emotions, thought processes, and the physical or psychological state of humans due to pressure from within or from outside a person, which can interfere with activities that can cause various diseases. Therefore, a tool is made to detect stress levels so that a person can monitor their condition and prevent stress from getting more severe and detrimental to the health of the body and mind. The stress level detection tool is designed using a galvanic skin response (GSR) sensor to detect skin response through a person's sweat glands and a heart rate sensor to detect heart rate. Furthermore, the reading results will be processed by microcontroller and then the stress level decision will be made using the fuzzy logic method and will be classified into Relax, Anxiety, Calm, and Stress. Based on the test results, the GSR parameter has the highest accuracy of 99.78%, and the heart rate parameter has the highest accuracy of 99.63%
Diabetes detection and prediction through a multimodal artificial intelligence framework
Get PDFDiabetes detection and prediction are crucial in modern healthcare, requiring advanced methodologies and comprehensive data analysis. This study aims to review the application of multi-parameters and artificial intelligence (AI) techniques in diabetes assessment, identify existing research limitations and gaps, and propose a novel multimodal framework for enhanced detection and prediction. The research objectives include evaluating current AI methodologies, analyzing multi-parameter integration, and addressing challenges in early detection and model evaluation. The study utilizes a systematic review approach, analyzing recent literature on AI-based diabetes detection and prediction, focusing on diverse data sources and machine learning (ML) techniques. Findings reveal a significant lack of integration of diverse data sources, limited focus on early detection strategies, and challenges in model evaluation. The study concludes with a proposed innovative framework for more accurate and personalized diabetes detection, contributing to the advancement of diabetes research and highlighting the potential of AI-driven healthcare interventions. This research underscores the importance of comprehensive data integration and robust evaluation methods in enhancing diabetes detection and prediction
Machine learning-based emotions recognition model using peripheral signals
Get PDFThis work proposes a system for emotion recognition using four peripheral signals electromyography, galvanic skin response, blood volume pulse, and respiration. Peripheral signals cannot be modified, unlike other expression like voice and facial expression. The proposed method is applied to the DEAP datasets to verify the accuracy of emotion recognition. The proposed model focuses on accuracy and F1-score. DEAP dataset has more signals but only thirty-seven features from four peripheral signals were extracted for each trail and each video. On the DEAP datasets, the implementation found that the classification accuracy for arousal, valence, liking, and dominance was, respectively, 80%, 75%, 71%, and 78%. For two classes of problems, the corresponding F1-scores for arousal, valence, liking, and dominance are 0.50, 0.49, 0.47, and 0.47. The proposed model was implemented in MATLAB R2017a
Linguistic feature selection for personality trait identification from textual data
Get PDFPersonality identification is a common and central problem in text processing. Sensing personality is helpful for various purposes; for example, estimating users' personalities before providing them with any service is necessary. Individuality is essential in a person's nature in every outlook, for instance, in text writing. But, this remains a core challenge because of the low accuracy achieved. The proposed study solves this problem and presents a big five trait identification technique from text data, which applies a feature selection method to increase accuracy. This technique is called linguistic feature selection for personality trait identification (LFSPTI). This technique first finds features based on mutual information (MI), F-statistic, principal component analysis (PCA), and chi-square, then uses the genetic algorithm (GA) to select high-ranked features from all feature subsets. These four parameters provide various forms of the dataset. The experimental results exhibit that the LFSPTI method enhances the classification accuracy against the best of the competing methods by 1.18%, 0.83%, 1.61%., 1.15%, 1.82%, and 1.39% for extraversion (EXT), neuroticism (NEU), agreeableness (AGR), conscientiousness (CON), openness (OPN), and mean overall personality traits, respectively