Indonesian Journal of Electrical Engineering and Computer Science
Not a member yet
9109 research outputs found
Sort by
A review of the impacts of linked open data on cross-domain recommender systems for individual and groups
Get PDFAs users' viewpoints on information searching change from information seeking to information receiving, new search paradigms are continuously emerging. Utilizing a recommender system (RS) is one of the modern ways to get information. The RS has succeeded in various traditional domains, including tourism, health, and books. However, some scenarios are more suitable to recommend to a group of users than an individual, such as listening to music at the same place and group traveling. The limited and incomplete number of user-item ratings triggers the challenges of the group and individual RSs. The data sparsity problem emerges because of this incompleteness. The quality of recommendations offered to individuals and groups suffers when there is data sparsity. Using knowledge gained from a source domain, cross-domain RSs can enhance recommendations in target domain. Cross-domain and linked open data approaches are two ways to increase recommendation systems' performance. The impacts of the two aforementioned approaches on individual and group RSs have been discussed. Furthermore, we highlighted various domains employed in cross-domain RSs for individuals and groups, examined diverse methodologies and algorithms, outlined current issues, and suggested future directions for cross-domain RSs research for groups leveraging linked open data technology
A novel deep learning based spatial delay feature aware encoder decoder module for enhanced CSI feedback in massive MIMO
Get PDFThe algorithm presented in this study addresses the challenge of reconstructing downlink channel state information (CSI) in massive multiple input multiple output (MIMO) systems with a focus on enhancing efficiency and accuracy. It begins by acquiring both downlink and uplink CSI data alongside other critical parameters such as the number of iterations and convolutional filter specifications. The process initiates with the vectorization of downlink CSI data followed by compression through a fully connected layer, effectively reducing dimensionality to manage computational complexity. The iterative reconstruction phase then unfolds, where each iteration updates an intermediary variable using a refined formula that incorporates the compressed CSI representation and correction factors. This iterative refinement aims to progressively enhance the accuracy of the reconstructed CSI. A pivotal aspect of the algorithm involves an optimized Encoder-Decoder framework designed to handle spatial-delay features inherent in MIMO systems. This framework employs thresholding operations to eliminate insignificant features, ensuring that the reconstructed CSI accurately reflects the crucial aspects of the channel. Simultaneously, an information module utilizes uplink CSI data to adjust weights during reconstruction, thereby further refining the accuracy of the downlink CSI estimation
A simulation-based investigation into the bidirectional charge and discharge dynamics in lead-acid batteries
Get PDFThis paper presents a comprehensive simulation-based investigation into the bidirectional charge and discharge dynamics of lead-acid batteries within electric vehicles (EVs) and energy storage systems (ESS). Utilizing a bidirectional DC-DC converter (BDC) integrated with a lead-acid battery, the study explores the performance of these batteries through various charging and discharging scenarios. The simulation model, implemented using MATLAB, assesses the impact of charging strategies on battery behavior, focusing on key metrics such as state of charge (SOC), energy performance, and charging rates. The results reveal that lead-acid batteries, when paired with appropriate charging infrastructure and strategies, demonstrate enhanced performance and reliability in both EV and ESS applications. The study highlights the significant role of BDC topology in facilitating efficient energy transfer and optimizing battery usage. The findings underscore the potential for improved performance and widespread adoption of bidirectional converters in sustainable energy solution
Segmentation and classification of plant leaf disease using advanced deep learning approach and ensemble classifier
Get PDFAn essential component of maintaining global food production is plants. On other hand, a number of plant diseases can threaten agricultural output and cause large losses if left unchecked. Agricultural specialists and botanists physically track plant diseases in a labor-intensive, error-prone manner using a conventional method. AI can give evaluations that are quicker and more accurate than those made using conventional approaches by automating the identification and analysis of diseases. This technical development presents a viable way to lessen crop losses and lessen the severity of infections. As a result, we describe an ensemble machine learning strategy for plant disease classification in this study that is enabled by deep learning. Data augmentation is done in the first part of the study, and in the second step, we provide a modified Mask R-CNN model for plant leaf segmentation. Afterwards, a model to extract the deep features based on CNN is shown. Lastly, the ensemble classifier is built using support vector machine classifier (SVM), random forest (RF), and decision tree (DT) with the aid of majority voting. The suggested method's effectiveness is tested on plant village, apple, maize, and rice, yielding overall accuracy values of 99.45%, 96.30%, 96.85%, and 98.25%, in that order
Design and implementation of smart farming prototype with renewable energy and IoT
Get PDFIndonesia faces food security challenges in several regions, and the adoption of advanced technologies such as artificial intelligence (AI), internet of thing (IoT), and renewable energy in the agricultural sector has not been optimal. This research aims to develop an integrated smart farming system, including monitoring, controlling, and prediction features based on renewable energy to support national food security, especially for chili plants. The method used in the research is an experiment, starting from analysis, design, manufacture, and testing. The result of the research is a smart farming prototype that has been tested with experts, partners and farmers. The results of expert testing obtained that the monitoring feature, in this case the accuracy is 4.36 out of 5 for all sensors, as well as the controlling and prediction features have met technical, functional, and practical needs. The results of the usability evaluation using the system usability scale (SUS) method involving partners and farmers obtained an average SUS score of 73.125. This result is categorized as an excellent rating and can be given a grade B and the acceptance range is high. So, from this study it can be concluded that the smart farming prototype can be used by chili farmers
Development of ResNet-18 architecture to lesion identification in breast ultrasound images
Get PDFBreast ultrasound (USG) is widely used for early breast cancer detection, but challenges such as noise, low contrast, and resolution limitations hinder accurate lesion identification. This study proposes a modified residual network-18 (ResNet-18) architecture for breast lesion segmentation, aimed at improving detection accuracy. The methodology involves preprocessing steps including red green blue (RGB) to Grayscale conversion, contrast stretching, and median filtering to enhance image quality. The modified ResNet-18 model introduces additional convolutional layers to refine feature extraction. The proposed model was trained and validated on 30 breast ultrasound images, with evaluation metrics including accuracy, sensitivity, and specificity. Experimental results indicate that the modified architecture outperforms the baseline model, achieving an average accuracy of 0.97093, sensitivity of 0.90056, and specificity of 0.97705. Validation by a radiology specialist confirms the model’s clinical relevance. These findings suggest that the enhanced ResNet-18 model has the potential to assist radiologists in more accurately identifying breast lesions. Future research should focus on expanding the dataset, integrating multi-modal imaging, and optimizing model generalizability for real-time clinical applications. The study contributes to advancing artificial intelligence (AI)-driven breast cancer diagnostics, supporting early detection, and improving patient outcomes
BdRegionText: resource creation and evaluation for Bangla regional text classification with machine learning
Get PDFRegional text analysis acknowledges the cultural diversity encompassed by a language. It offers insights into the authentic ways people communicate, promoting cultural awareness and genuineness in communication. This research paper delves into the classification of Bangla regional text using machine learning (ML) algorithms. Consequently, this study compiles a dataset comprising 2,573 sample texts in four distinct regional Bangla dialects (Chittagong, Rangpur, Barishal, and Noakhali). We focused on these dialects because they were more readily available on the internet than others. The primary objective is to identify synthesized Bangla text and assign appropriate categories. The categorization process focuses on a regional language authored by Bengali individuals, aiming to ascertain its authenticity and using ML techniques named decision tree (DT), stochastic gradient descent (SGD), support vector machine (SVM), and random forest (RF) to check how well categorization worked and also handled the issue of slight imbalance in the dataset. As there is limited prior research in this domain, we compare our work with the existing studies available, and we have employed various popular feature extraction techniques for text classification in natural language processing (NLP), specifically TF-IDF, CountVectorizer, and bag of words (BoW). Our comparative analysis indicates that an aggregation of term frequency–inverse document frequency (TF-IDF) and CountVectorizer outperforms BoW in terms of performance. Among the ML techniques we applied, the RF algorithm yielded the utmost accuracy of 79.15% and a mean accuracy of 79.47%
Advanced cloud security framework based on zero trust architecture and adaptive deep learning for next-generation systems
Get PDFStatic rule-based models and cloud access security brokers (CASBs) — traditional cloud security frameworks— can no longer effectively mitigate modern and evolving cyber threats. Two such examples include signature-based detection methods which lack real-time versatility and are ineffective against advanced persistent threats or zero-day threats. In this paper, we introduce an adaptive zero trust framework (AZTF) based on the integration of zero trust architecture (ZTA) and adaptive deep learning (ADL) approach to dynamically evaluate threats and risks being targeted on cloud environments. It continually monitors access attempts using DL models for real-time anomaly detection. Nine synthetic datasets were generated and used in the experiment in two security domains: network traffic and access pattern. The proposed system reached 96% detection accuracy, 52% improvements in response time, and 12% resource consumption optimization compared to traditional ZTA-based security models. The results highlight the power of using a combination of continuous authentication with artificial intelligence (AI)-powered dynamic security policy application to strengthen the resilience of cloud security. Future research will focus on federated learning integration, multi-cloud security applications, and explainable AI for increased transparency of models
A feeling classification model in a blood draw situation using power spectrum density and a random forest algorithm
Get PDFFeelings and expressions such as pain, anxiety, and excitement can occur while getting blood drawn. These are the physical symptoms that can occur in some patients. A medical provider cannot know pain or anxiety symptoms, which could cause harm to the patients throughout the procedure. However, electroencephalography (EEG) changes, such as Delta, Theta, Alpha, Beta, and Gamma, are essential to identify the patient’s feelings. These can assist in decreasing danger during the procedures. Therefore, this research aims to investigate the patterns in the power spectrum density (PSD) form to classify two feeling states during blood drawing: normal and anxious feelings. This research focused on alpha, beta, and gamma of the PSD. Thus, a method was designed based on the changing values of alpha, beta, and gamma. Each PSD of three waves was derived at 56 minutes. The pattern from this dataset was applied to classify feeling expressions using a random forest (RF) algorithm. This algorithm was used to create a feeling classification model (FCM). The accuracy of the FCM in classifying feeling differences between normal and anxious feelings was 100%. Thus, this proves that the FCM is highly efficient
Multi-camera multi-person tracking with DeepSORT and MySQL
Get PDFMulti-camera multi-object tracking refers to the process of simultaneously tracking numerous objects using a network of connected cameras. Constructing an accurate depiction of an object’s movements requires the analysis of video data from many camera feeds, detection of items of interest, and their association across various camera perspectives. The objective is to accurately estimate the trajectories of the objects as they navigate through a monitored area. It has several uses, including surveillance, robotics, self-driving cars, and augmented reality. The current version of an object tracking algorithm, DeepSORT, doesn’t account for errors caused by occlusion or implementation of multiple cameras. In this paper, DeepSORT has been extended by introducing new states to improve the tracking performance in scenarios where objects are occluded in the presence of multiple cameras. The communication of track information across multiple cameras is achieved with the help of a database. The suggested system performs better in situations where objects are occluded, whether due to object occlusions or person occlusions