Articles

Access the latest knowledge in applied science, electrical engineering, computer science and information technology, education, and health.

Advanced Search
Filter Icon

Filters article

Years

FAQ Arrow
Single Year
Range Year
0
0

Source Title

FAQ Arrow

Authors

FAQ Arrow

31,042 Article Results

Flow-guided long short-term memory with adaptive directional learning for robust distributed denial of service attack detection in software-defined networking

10.11591/ijece.v15i6.pp5484-5496
Huda Mohammed Ibadi , Asghar A. Asgharian Sardroud
A software-defined networking (SDN) architecture is designed to improve network agility by decoupling the control and data planes, but while much more flexible, also makes networks more vulnerable to threats, such as distributed denial of service (DDoS) attacks. In this study we present a novel detection model, the flow-guided long short-term memory (LSTM) network with adaptive directional learning (ADL), for the mitigation of DDoS attacks in software defined networking (SDN) environments. While the methodology is based on a flow direction algorithm (FDA), which analyzes traffic patterns and detects anomalies from directional flow behavior. The proposed method integrates FDA in LSTM-based threat detection frameworks within internet of things (IoT) networks, thereby yielding enhanced detection accuracy, as well as a real-time security threat response. The experimental evaluation on two benchmark datasets, namely the InSDN dataset and a real-time dataset utilizing a Mininet and POX controller setup, shows that a detection rate of 99.85% and 99.72%, respectively, thereby showcasing the proposed model’s ability to differentiate between legitimate and malicious network traffic.
Volume: 15
Issue: 6
Page: 5484-5496
Publish at: 2025-12-01

Analysis and implementation of computation offloading in fog architecture

10.11591/ijra.v14i4.pp479-492
Prince Gupta , Rajeev Sharma , Sachi Gupta , Adesh Kumar
The fast expansion of connected devices has led to an unparalleled increase in data across sectors like industrial automation, social media, environmental monitoring, and life sciences. The processing of this data presents difficulties owing to its magnitude, temporal urgency, and security stipulations. Computation offloading has arisen as a viable alternative, allowing resource-constrained devices to assign demanding work to more robust platforms, thus improving responsiveness and efficiency. This paper examines decision-making strategies for computing offloading by assessing various algorithms, including a deep neural network with deep reinforcement learning (DNN-DRL), coordinate descent (baseline), AdaBoost, and K-nearest neighbor (KNN). The performance evaluation centers on three primary metrics: system accuracy, training duration, and latency. The computation offloading mitigates these issues by transferring intricate workloads from resource-limited devices to more proficient platforms, thus enhancing efficiency and responsiveness. The evaluation examines accuracy, training duration, and latency as key parameters. The results indicate that KNN attains maximum accuracy and minimal latency, AdaBoost provides a robust balance despite increased training costs, and the baseline underperforms in both efficiency and responsiveness. These findings underscore the trade-offs between computational expense, precision, and real-time application, providing insights for forthcoming IoT and edge-computing systems.
Volume: 14
Issue: 4
Page: 479-492
Publish at: 2025-12-01

Fit to-organization amplifies unethical pro-organizational behavior

10.11591/ijaas.v14.i4.pp1050-1060
Hany Azza Umama , Sri Murni Setyawati , Siti Zulaikha Wulandari
The study of unethical pro-organizational behavior (UPB) has seen a sharp rise in recent years, with research results explaining that a potential cause of UPB is organizational identification (OI). However, there are inconsistencies in the findings regarding the effect of OI on UPB in the workplace. This study seeks to test the direct effect of OI on UPB, and to explore the mediating role of fit to-organization (Fit-O). To evaluate data collected from employees of micro and small enterprises, this study utilized structural equation modeling-partial least squares (SEM-PLS) analysis, along with the variance accounted for (VAF) method to test the mediation between OI and UPB. The findings confirm that the fit or compatibility of an individual with the organization can strengthen the effect of OI on UPB. The intervening role of Fit-O in the OI-UPB relationship is a crucial theoretical contribution. This research also implies that organizations must balance increasing OI with strong ethical standards to mitigate UPB.
Volume: 14
Issue: 4
Page: 1050-1060
Publish at: 2025-12-01

Large language models for pattern recognition in text data

10.11591/ijai.v14.i6.pp5311-5332
Aknur Kosayakova , Kurmashev Ildar , Luigi La Spada , Nida Zeeshan , Makhabbat Bakyt , Moldamurat Khuralay , Omirzak Abdirashev
Large language models (LLMs) are widely deployed in settings where both reliability and efficiency matter. We present a calibrated, seed‑robust empirical comparison of an encoder fine‑tuned model (bidirectional encoder representations from transformers (BERT)‑base) and a decoder in‑context model (generative pre-trained transformer (GPT)‑2 small) across Stanford question answering dataset v2.0 (SQuAD v2.0) and general language understanding evaluation (GLUE)-multi-genre natural language inference (MNLI), Stanford sentiment treebank 2 (SST‑2). Beyond accuracy, we assess reliability (expected calibration error with reliability diagrams and confidence–coverage analysis) and efficiency (latency, memory, throughput) under matched conditions and three fixed seeds. BERT‑base yields higher accuracy and lower calibration error, while GPT‑2 narrows gaps under few‑shot prompting but remains more sensitive to prompt design and context length. Efficiency benchmarks show that decoder‑only prompting incurs near‑linear latency/memory growth with k‑shot exemplars, whereas fine‑tuned encoders maintain stable per‑example cost. These findings offer practical guidance on when to prefer fine‑tuning versus prompting and demonstrate that reliability must be evaluated alongside accuracy for risk‑aware deployment.
Volume: 14
Issue: 6
Page: 5311-5332
Publish at: 2025-12-01

Review of NLP in EMR: abbreviation, diagnosis, and ICD classification

10.11591/ijict.v14i3.pp881-891
Nurul Anis Balqis Iqbal Basheer , Sharifalillah Nordin , Sazzli Shahlan Kasim , Azliza Mohd Ali , Nurzeatul Hamimah Abdul Hamid
This review explores state-of-the-art natural language processing (NLP) methods applied to electronic medical records (EMRs) for key tasks such as expanding medical abbreviations, automated diagnosis generation, international classification of diseases (ICD) classification, and explaining model outcomes. With the growing digitization of healthcare data, the complexity of EMR analysis presents a significant challenge for accurate and interpretable results. This paper evaluates various methodologies, highlighting their strengths, limitations, and potential for improving clinical decision-making. Special attention is given to abbreviation expansion as a crucial step for disambiguating terms in the clinical text, followed by an exploration of auto-diagnosis models and ICD code assignment techniques. Finally, interpretability methods like integrated gradients and attention-based approaches are reviewed to understand model predictions and their applicability in healthcare. This review aims to provide a comprehensive guide for researchers and practitioners interested in leveraging NLP for clinical text analysis.
Volume: 14
Issue: 3
Page: 881-891
Publish at: 2025-12-01

A merchant analytics framework for revenue forecasting and financial stress detection using transaction data

10.11591/ijai.v14.i6.pp4848-4864
Yara Harb , Wissam Baaklini , Nadine Abbas , Seifedine Kadry
By processing payments and providing specialized financial services, acquiring banks are essential for merchants’ operations. To forecast 30-day revenue trajectories, identify seasonal demand patterns, and identify early indicators of financial stress, this paper presents a scalable merchant analytics framework that benefits from transactional data. The framework captures multi-level seasonalities using Prophet time series model, allowing dynamic product offerings like revenue-based loans. Proactive risk management is supported offerings like revenue-based loans. Proactive risk management is supported. by a new stress-flagging mechanism that identifies merchants at risk based on deviations in revenue trends. The framework achieved a median 30-day mean absolute percentage error (MAPE) of 56.51% after the validation on a dataset with 130,350 transactions from 460 merchants in a volatile economic environment. The model demonstrated significant practical utility in identifying financial distress and segmenting merchant behavior, despite its moderate predictive precision, which is common challenge in high-variance merchant datasets. Model outputs are converted into decision-support visualizations along with an interactive dashboard.
Volume: 14
Issue: 6
Page: 4848-4864
Publish at: 2025-12-01

Predicting the severity of road traffic accidents Morocco: a supervised machine learning approach

10.11591/ijai.v14.i6.pp4461-4473
Halima Drissi Touzani , Sanaa Faquir , Ali Yahyaouy
Early prediction of road accidents fatality and injuries severity is one of the important subjects to road safety emphasizing the critical need to prevent serious consequences to reduce injuries and fatalities. This study uses real road accidents data set in Morocco. It represents the intersection between road safety and data science, aiming to employ machine learning techniques to provide valuable insights in accident’s severity prevention. The purpose of this paper is to study road accidents data in the country and combine results from statistical methods, spatial analysis, and machine learning models to determine which factors will mostly contribute to increase the accident’ severity in the country. A comparison of results obtained was also conducted in this paper using different metrics to evaluate the effectiveness of each method and determine the most important factors that contribute to increase the fatality or injuries severity in the specific context of accidents. The best prediction model was then injected into a proposed algorithm where more intelligent techniques are included to be implemented in a car engine to perform an early detection of severe accidents and therefore preventing crashes from happening.
Volume: 14
Issue: 6
Page: 4461-4473
Publish at: 2025-12-01

Metaheuristic optimization for sarcasm detection in social media with embedding and padding techniques

10.11591/ijai.v14.i6.pp5027-5037
Geeta Sahu , Manoj Hudnunkar
Sarcasm is a sophisticated mode of expression that allows speakers to express their opinions subtly. Stakeholders provide unstructured messages with extended phrases, making it difficult for computers and people to understand. This research aims to develop a sarcasm detection method to identify words in phrases as sarcastic or non-sarcastic from text, utilizing natural language processing appliances. The first step is pre-processing, when the padding and embedding are performed. Zero padding and end padding are used for the padding. At the same time, different embedding techniques, such as word2vec, Glove, and BERT, are used. Following pre-processing, the features are extracted from the pre-processed data, including "information gain, chi-square, mutual information, and symmetrical uncertainty-based features." Then, a hybrid optimization technique known as clan-updated grey wolf optimization (CU-GWO) is used for optimized features and weight selection. An ensemble technique was applied to extract optimal features. The classifiers in the proposed suggested ensemble technique with deep convolution neural network (DCNN). DCNN offers fine weight tuning and detection results.The performance analysis and its impact on the proposed model for sarcasm detection are classified with good accuracy into sarcastic and non-sarcastic categories. The results are also compared with against those of the GloVe and BERT techniques.
Volume: 14
Issue: 6
Page: 5027-5037
Publish at: 2025-12-01

Enhancing academic conferences with AI: defining the role of the human AI editor

10.11591/ijai.v14.i6.pp4484-4493
Esteban Galan-Cubillo , Emilio Saez-Soro
Academic conferences serve as key platforms for knowledge exchange, yet they face challenges in managing large volumes of content efficiently while maintaining academic rigor. To address these challenges, this study introduces and evaluates the "AI editor": a novel human expert role who, using tools like ChatGPT, supervises, refines, and structures artificial intelligence (AI)-generated content in real time. Through a mixed-methods approach, we examine the role of AI in enhancing content creation and engagement. This approach included the experimental deployment of the AI editor in three sustainability-focused European academic conferences (in Spain and UK) and formative workshops with 127 university students from the same countries. While AI-assisted tools improve efficiency, concerns persist regarding traceability, reliability, and ethical oversight. Our findings indicate that AI by itself cannot guarantee scholarly integrity; continuous human oversight is indispensable. The AI editor ensures coherence, quality control, and compliance with academic standards, addressing a critical gap in AI adoption within research environments. This study contributes to the discourse on responsible AI use in academia by proposing a structured framework for its integration into conferences, balancing automation with human oversight. Moreover, it highlights the growing need for digital intelligence that enables researchers to interact ethically and effectively with AI and other digital technologies, fostering responsible and informed academic innovation.
Volume: 14
Issue: 6
Page: 4484-4493
Publish at: 2025-12-01

A smart grid fault detection using neuro-fuzzy deep learning algorithm

10.11591/ijai.v14.i6.pp5096-5105
Etienne Francois Mouckomey , Jacques Bikai , Camille Franklin Mbey , Alexandre Teplaira Boum , Felix Ghislain Yem Souhe , Vinny Junior Foba Kakeu
This paper proposes a novel data analysis framework that integrates deep learning with a binary neuro-fuzzy algorithm to address the problem of fault localization in smart power grids. In the first stage, a long short-term memory (LSTM) network is employed to train data samples collected from smart meters. The resulting learned features are subsequently utilized by an adaptive neuro-fuzzy inference system (ANFIS) for accurate fault detection and classification. Through this intelligent hybrid approach, multi-phase faults can be efficiently identified using a limited amount of data. The proposed method distinguishes itself by its capacity to rapidly train and test large datasets while maintaining high computational efficiency. To evaluate the performance of the model, an advanced simulation of the IEEE 123-node test feeder is conducted. The robustness and effectiveness of the proposed framework are validated using multiple performance metrics, including precision, recall, accuracy, F1-score, computational complexity, and the ROC curve. The results demonstrate that the proposed deep learning–based model significantly outperforms existing approaches in the literature, achieving a fault detection and classification precision of 99.99%.
Volume: 14
Issue: 6
Page: 5096-5105
Publish at: 2025-12-01

Fine-tuning multilingual transformers for Hinglish sentiment analysis: a comparative evaluation with BiLSTM

10.11591/ijai.v14.i6.pp4684-4693
Jyoti S. Verma , Jaimin N. Undavia
Growing trend of code-mixing in languages, in the form of Hinglish, greatly tests the skills of conventional sentiment analysis tools. The research contributes a fine-tuned multilingual transformer model built exclusively for classifying sentiment of Hinglish customer reviews. Drawing from pre trained BERT-base-multilingual-case architecture, the model gets transformed with the process of fine-tuning the same on synthetically prepared and balanced dataset simulating positive, negative, and neutral sentiments. Sophisticated methods like focal loss for addressing the class imbalance and mixed precision training for maximization of computational effectiveness are embedded within the training process. Experimental results suggest that the proposed method significantly captures the fine-grained linguistic patterns of code-mixed text, improving sentiment classification accuracy. The results show promising potential for enhancing customer feedback analysis in e-commerce, social media monitoring, and customer support use cases, where it is crucial to comprehend the sentiment behind code-mixed reviews.
Volume: 14
Issue: 6
Page: 4684-4693
Publish at: 2025-12-01

Classification of single origin Indonesian coffee beans using convolutional neural network

10.11591/ijai.v14.i6.pp5140-5156
Achmad Pratama Rifai , Wangi Pandan Sari , Haidar Rabbani , Tari Hardiani Safitri , Makbul Hajad , Edi Sutoyo , Huu-Tho Nguyen
This research aims to develop a coffee bean type detection model using convolutional neural networks (CNN), leveraging a dataset of 14,525 images from 116 types of Indonesian coffee beans. Pre-processing steps including resizing, rescaling, and augmentation were applied to improve the dataset quality. The dataset was split into training, validation, and testing sets with proportions of 80%, 10%, and 10%, respectively. Two model development approaches were used: transfer learning with Inception V3 in two scenarios and a model built from scratch. The transfer learning Inception V3 model in scenario 1 achieved the best performance, with a test accuracy of 0.87 and optimal evaluation metrics across precision, recall, and F1-score. This model was fine-tuned using pretrained weights, allowing it to adapt effectively to the coffee bean dataset. The results highlight that transfer learning, especially with Inception V3, provides a robust method for classifying coffee beans, offering potential applications in the coffee industry for improving classification efficiency and accuracy. The study demonstrates how deep learning can enhance the objectivity and precision of coffee bean classification, contributing to greater consistency in product sorting and quality assessment.
Volume: 14
Issue: 6
Page: 5140-5156
Publish at: 2025-12-01

Low-speed scalar control of induction motor by fuzzy logic

10.11591/ijai.v14.i6.pp4623-4635
Alfonso Alejandro Sevilla-Hidalgo , Rossy Uscamaita-Quispetupa , Julio Cesar Herrera-Levano , Limberg Walter Utrilla Mego , Roger Jesus Coaquira-Castillo
Efforts have continually been directed toward optimizing processes in various fields, and the application in induction motors is no exception. Scalar control V/f offers a straightforward method to regulate the speed of a three-phase induction motor (TIM). However, it faces challenges at low speeds or proportionally at low frequencies, often failing to operate below 20% of its rated speed. This control typically pairs with a PI controller (PIC) for closed loop speed regulation, but its limited control range hinders performance at low speeds. Although intelligent methods have been developed to improve scalar V/f control, attention is often focused on high speeds, while control at low speeds is overlooked. This paper presents the simulation of a fuzzy controller (FC) with a Mamdani-type structure designed to achieve effective low-speed control of a TIM using the V/f scalar control technique. The results not only show improvements in overshoot and settling time but also reveal that the FC can control speeds as low as 6.06% of the rated speed, and it ensures a starting current below the designed motor current under load. Comparative analysis indicates that the FC outperforms the PIC in low-speed control, and it provides an optimal inrush current across different low speeds.
Volume: 14
Issue: 6
Page: 4623-4635
Publish at: 2025-12-01

Spam social media profile detection using hybrid positive unlabelled learning

10.11591/ijai.v14.i6.pp4838-4847
Nidhi A. Patel , Nirali Nanavati
Online social networks (OSNs) are a communication medium of social interaction for people, where social activities, entertainment, business oriented activities, and information are exchanged. It creates an environment with worldwide connectivity where groups of individuals may discuss their interests and activities on social media platforms. Billions of people routinely interact with social content, opinion sharing, recommendations, networking, scouting, social campaigns, alerting on OSNs. The increase in popularity of OSNs creates new challenges and perspectives to the researchers of social networks, which is of interest in various fields. One of the most popular networking platforms for microblogging is X (formerly Twitter). Millions of spam accounts have inundated the X network, which could damage normal users' security and privacy. Hence, the research in this filed has become essential for enhancing real users' protection and identifying spam profiles. In this manuscript, we propose hybrid approach based on semi-supervised learning to detect the spam profiles. The proposed work is based on the positive and unlabeled (PU) learning algorithm, which learns from an unlabeled dataset and a small number of positive instances. Simulation results demonstrate that our approach outperformed existing PU learning approach by 17.39% and 17.51% improvement respectively in spam detection rate on X and Instagram datasets.
Volume: 14
Issue: 6
Page: 4838-4847
Publish at: 2025-12-01

Deep neural network solutions to Newell-Whitehead-Segel equations

10.11591/ijai.v14.i6.pp5172-5182
Soumaya Nouna , Ilyas Tammouch , Assia Nouna , Mohamed Mansouri
In this work, we use the deep neural network (DNN) approach called NeuroDiffEq, and the unified finite difference exponential approach for obtaining the approximated and exact solutions of Newell-Whitehead-Segel systems that are essential for the biology of mathematics. A unified approach was used to generate several solutions for solitary waves of those systems. The approximated solutions for selected studies are explored using the NeuroDiffEq approach, which is the artificial neural networks (ANN) approach and is based upon trial approximate solution (TAS). The comparison between the obtained approximated solutions and the analytical solutions indicates that the applied method has proved an efficient as well as a highly successful approach to solving various types of the Newell-Whitehead-Segel equations.
Volume: 14
Issue: 6
Page: 5172-5182
Publish at: 2025-12-01
1 2 3
...
2069 2070
Show 153 of 2070

Discover Our Library

Embark on a journey through our expansive collection of articles and let curiosity lead your path to innovation.

Explore Now
Library 3D Ilustration