Indonesian Journal of Electrical Engineering and Computer Science
Not a member yet
9109 research outputs found
Sort by
A compact study on methodological insights on navigational systems in vehicular traffic system
Get PDFNavigation system has witnessed a significant inclusion of potential technological advancement in the area of vehicular traffic system. Since the last decade, there are various evolution of innovative techniques that has identified and addressed some serious problem towards vehicular navigation system. With a progress of time, artificial intelligence (AI) has evolved as contributory role model towards optimizing the performance of navigation system. However, still it is quite challenging to acquire a quick snapshot of overall stand of all such methodologies and its effectiveness. Hence, this paper presents a precise, compact, and highly crisp discussion of core taxonomies of methods towards improving navigation system. The paper also contributes towards highlighting their strength and weakness followed by updated research trend to understand the true picture. Finally, the paper contributes to highlight the critical trade-off and gaps
Benchmarking spectral handoff rate performance in cognitive wireless networks with real multi-user access
Get PDFCognitive radio (CR) has proven to be an excellent alternative to the problem of inefficient spectrum use in wireless networks. However, the vast majority of proposals found in the current literature are restricted to the access of a single secondary user (SU) to the network, and the few proposals with multiple access do not take into account the access of other primary users (PUs) during the opportunistic transmission of the SU. The objective of this work is to perform a comparative evaluation of the spectral handoff (SH) rate in cognitive wireless networks with multi-user access in an environment with other PUs interacting. To carry out this evaluation, four SH models with better performance were selected: deep learning (DL), feedback fuzzy analytic hierarchy process (FFAHP), simple additive weighting (SAW), and Naïve Bayes (NB), which were validated according to the metric of the number of total handoffs, under four scenarios given by the combination of the following parameters: low spectral availability, high spectral availability, active presence of others SUs, and passive presence of others SUs. The results show that each model performs well according to the scenario in which it is executed, suggesting an adaptive multi-model as a proposal
Vulnerability detection in smart contact using chaos optimization-based DL model
Get PDFThis research article introduces a deep learning (DL) for identifying vulnerabilities in the smart contracts, leveraging an optimized DL method. The proposed method, termed LogT BiLSTM, combines bidirectional long short-term memory (BiLSTM) with logistic chaos Tasmanian devil optimization (LogT) for enhancing detection of vulnerability. The evaluation of the suggested approach is conducted using publicly available datasets. Initially, preprocessing steps involve removing duplicate data and imputing missing data. Subsequently, the vulnerability detection process utilizes BiLSTM, with the optimization of the loss function achieved through LogT. Results indicate promising performance in identifying vulnerabilities in SC, highlighting the efficacy of the LogT-BiLSTM approach
Integrating swarm intelligence with deep learning for enhanced social media sentiment analysis
Get PDFUnderstanding user views on social media in the advent of internet content demands sentiment analysis. This study introduces a novel approach called particle swarm-accelerated model (PSAM), that integrates deep learning with long short-term memory (LSTM) with two hyper-parameters and swarm intelligence through particle swarm optimization (PSO). In the sentiment classification of YouTube movie reviews for “Pushpa 2,” the recommended approach classifies opinions as “positive,” “negative,” or “neutral,” with an accuracy score of 95.3%. The process involved utilizing YouTube API to collect user-genearted comments, followed by advanced preprocessing steps such as punctuation removal, stopword filtering, slang normalization, and emoji handling. PSO performs feature selection to boost the efficiency of classification systems. The PSAM model reaches superior outcome results compared to support vector machines (SVM), Naive Bayes, CNN, and random forest classifiers when evaluated based on F1-score and accuracy metrics. The proposed hybrid model demonstrates its ability to boost sentiment analysis in different social media platforms according to research findings
Boosting real-time vehicle detection in urban traffic using a novel multi-augmentation
Get PDFReal-time vehicle object detection in urban traffic is crucial for modern traffic management systems. This study focuses on improving the accuracy of vehicle identification and classification in heavy traffic during peak hours, with particular emphasis on challenges such as small object sizes and interference from light reflections. The use of multi-label images enables the simultaneous detection of various vehicle types within a single frame, providing more detailed information about traffic conditions. You only look once (YOLO) was chosen for its capability to perform real-time object detection with high accuracy. Multi-augmentation techniques were applied to enrich the training data, making the model more robust to varying lighting conditions, viewpoints, object occlusions, and issues related to small objects. YOLOv8n and YOLOv9t were selected for their speed and efficiency. Models without augmentation, 10 single-augmentation techniques, and 5 multi-augmentation techniques were tested. The results show that YOLOv8n with multiaugmentation (scaling, zoom in, brightness adjustment, color jitter, and noise injection) achieved the highest mAP50-95 score of 0.536, surpassing YOLOv8n with single-augmentation Blur, which had an mAP50-95 of 0.465, as well as YOLOv8n without augmentation, which scored 0.390. Multiaugmentation proved to significantly enhance YOLO’s performance
An efficient machine learning framework for optimizing hyperspectral data analysis in detecting adulterated honey
Get PDFHoney adulteration detection involves employing spectral data, often utilizing machine learning (ML) techniques, to identify the presence of impurities or additives in honey. This study aims to explore ML models through the collection of a hyperspectral honey dataset with limited samples and 128 features. Three distinct feature selection (FS) methods i.e., Boruta, repeated incremental pruning to produce error reduction (RIPPER), and gain ratio attribute evaluator (GRAE) are applied to extract important features for decision-making. Then, the feature-selected dataset is classified through four effective ML algorithms, such as support vector machine (SVM), random forest (RF), logistic regression (LR), and decision tree (DT). Accuracy, F1-score, Kappa Statistics, and Matthews correlation coefficient (MCC) are the performance metrics used to assess the results of ML algorithms. RIPPER FS technique gave the best results by improving its accuracy values from 79.05% (primary data) to 91.89% (augmented data) for the RF classifier model and 74.93% (primary data) to 91.89% (augmented data) for the DT classifier model. These detailed examinations of the experiments demonstrate that proper finetuning of the ML methods can play a vital role in optimizing hyperspectral data analysis for detecting adulteration levels in honey samples
Analysis of real-time multi-surveillance detection model using YOLO v5
Get PDFImplementation of this advanced nighttime monitoring system provides one of the basic requirements toward the creation of an intelligent urban environment. The nighttime effective monitoring is highly enabled due to seamless integration of multi-directional cameras working as advanced sensors enhancing security measures in smart cities. This paper addresses the mentioned issues directly by proposing the you only look once version 5 (YOLOv5) model dedicated to object detection. It is experimentally confirmed, based on the dataset results, that the mean average precision of YOLOv5 multi-scale (YOLOv5MS) reaches an impressive 88.7%. The results unmistakably confirm domination of the model and its good ability to work over a network of more than 50 security cameras under the high restrictions of our operation. The use of state-of-art nighttime surveillance systems is an important constituent element in the construction of smart urban environment. The smooth interaction between multiple-angle cameras, which work as perceptive sensors, substantially upgrades the functionality of nighttime surveillance and strengthens security practices for smart cities. The current work presented the YOLOv5 model specifically designed for the task of target detection, targeting these issues head-on. The empirical data obtained from the dataset point to an outstanding mean average precision (mAP) of 88.7% for YOLOv5MS. Such results clearly prove the superiority of the model and demonstrate its excellent performance in a network of more than 50 security cameras under our harsh operational conditions
Detecting pneumonia from chest X-rays using deep learning based neural networks: an hybrid approach
Get PDFPneumonia a disease which occurs when the alveoli (air sacs) in the lings fill with fluidlike substance it can be due to infectious agents like virus, bacteria especially in an environment with contaminated air is often considered as lethal disease because the deaths associated with is high. There are several factors which contribute to this disease like age as their immune systems are not fully developed making it easier to get attacked by infections, chronic health conditions like asthma or weak immune systems may worsen the situation. Machine learning (ML) algorithms have tend to perform better while images are given, however compared to them deep learning (DL) algorithms have shown good promising results especially when images are given as an input this is because they have upper hand in identifying key features and loss optimization makes them best suited for this tasks. The significance of this research is to make an extensive review on the pneumonia and early detecting pneumonia by utilizing DL based neural networks
Trends in machine learning for predicting personality disorder: a bibliometric analysis
Get PDFOver the last decade, research on artificial intelligence (AI) in the medical field has increased. However, unlike other disciplines, AI in personality disorders is still in the minority. For this reason, we conduct a map research using bibliometric and build a visualization map using VOSviewer in AI to predict personality disorders. We conducted a literature review using the systematic literature review (SLR) method, consisting of three stages: planning, implementation, and reporting. The evaluation involved 22 scientific articles on AI in predicting personality disorders indexed by Scopus Quartile Q1–Q4 from the Google Scholar database during the last five years, from 2018–2023. In the meantime, the results of bibliometric analysis have led to the discovery of information about the most productive publishers, the evolution of scientific articles, and the quantity of citations. In addition, VOSviewer’s visualization of the most frequently occurring terms in abstracts and titles has made it easier for researchers to find novel and infrequently studied subjects in AI on personality disorders
Adaptive mathematical modeling for predicting and analyzing malware
Get PDFIn this paper, we propose and investigate an improved mathematical model of malware propagation in network structures based on a modification of the well-known raw-immune-response susceptible-infected-recovered (SIR) model. For detailed numerical analysis, our study introduces the fourth-order Runge-Kutta method, which provides higher accuracy in determining fundamental parameters such as infection, recovery and immunity loss coefficients of network nodes. The obtained simulation results demonstrate that the peak of the epidemic occurs when 34.7% of all nodes are infected, with a peak after 32.5-time units. The main contribution of this work is the in-depth understanding and quantification of cyber threats, which emphasizes the importance of prompt response, regular system software updates, and continuous monitoring of network activity. This research makes a significant contribution to cybersecurity applications by providing quantitative tools and strategies to help strengthen network defenses against malicious attacks. The identified patterns and their numerical interpretation can be integrated into processes for optimizing measures to prevent the widespread spread of malware, thereby enhancing the overall security and stability of networked systems