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

31,456 Article Results

Enhancing touchless smart locker systems through advanced facial recognition technology: a convolutional neural network model approach

10.11591/ijai.v14.i4.pp3262-3273
Abdul Haris Rangkuti , Evawaty Tanuar , Febriant Yapson , Felix Octavio Sijoatmodjo , Varyl Hasbi Athala
As the world recovers from COVID-19, demand for contactless systems is increasing, promising safety and convenience. Touchless technology, particularly public locker security systems that use facial recognition and hand detection, is advancing rapidly. The system minimizes physical contact, increasing user safety. It uses advanced models such as multi-task cascaded convolutional networks (MTCNN) and RetinaFace, FaceNet512, ArcFace, and visual geometry group (VGG)-Face for face detection and recognition, with a combination of RetinaFace, ArcFace, and L2 norm Euclidean or cosine as the most effective distance metric method, where the accuracy reaches 96 and 90%. 'Yourvault', an application demonstrating this efficient security feature, provides notifications for mask detection, facial authenticity and locker status, offering a solution to the problem of convenience and security of public spaces. Future research could investigate the impact of photo age on facial recognition accuracy, potentially making touchless systems more efficient. In general, the application of this technology is an important step towards a safer and more comfortable world after the pandemic. This model approach can be followed up with more optimal facial recognition.
Volume: 14
Issue: 4
Page: 3262-3273
Publish at: 2025-08-01

Synthesizing strategies and innovations in combating land degradation: a global perspective on sustainability and resilience

10.11591/ijai.v14.i4.pp3133-3142
Gangamma Hediyalad , Ashoka Kukkuvada , Govardhan Hegde Kota
This paper presents a comprehensive examination of land degradation, a critical environmental challenge with far-reaching implications for agricultural productivity, ecosystem sustainability, and socio-economic stability worldwide. With the backdrop of escalating human population pressures and the exacerbating impact of climate change. It delves into the causes and consequences of soil erosion, desertification, salinization, and biodiversity loss, highlighting the interplay between natural processes and anthropogenic activities. Through a detailed review of literature spanning various remediation technologies, conservation practices, and policy frameworks, the paper critically assesses the effectiveness of current land management approaches, including the utilization of biosurfactants, remote sensing technologies, and agroforestry systems. Furthermore, it identifies significant research gaps and future directions, emphasizing the need for quantitative assessments, exploration of socio-economic impacts, and evaluation of restoration techniques. By offering evidence-based recommendations for policymakers and practitioners, this paper contributes to the global dialogue on sustainable land management and aims to catalyze action towards halting the advance of land degradation, ensuring food security, and preserving biodiversity for future generations. This work not only advances our understanding of land degradation challenges but also outlines a path forward for research, policy, and practice in the pursuit of environmental sustainability and resilience.
Volume: 14
Issue: 4
Page: 3133-3142
Publish at: 2025-08-01

Human detection in CCTV screenshot using fine-tuning VGG-19

10.11591/ijict.v14i2.pp645-652
Firdaus Angga Dewangga , Abba Suganda Girsang
Closed-circuit television (CCTV) systems have generated a vast amount of visual data crucial for security and surveillance purposes. Effectively categorizing security level types is vital for maintaining asset security effectively. This study proposes a practical approach for classifying CCTV screenshot images using visual geometry group (VGG-19) transfer learning, a convolutional neural network (CNN) classification model that works really well in image classification. The task in classification compromise of categorizing screenshots into two classes: “humans present” and “no humans present.” Fine-tuning VGG-19 model attained 98% training accuracy, 98% validation accuracy, and 85% test accuracy for this classification. To evaluate its performance, we compared fine-tuning VGG-19 model with another method. The VGG-19-based fine-tuning model demonstrates effectiveness in handling image screenshots, presenting a valuable tool for CCTV image classification and contributing to the enhancement of asset security strategies.
Volume: 14
Issue: 2
Page: 645-652
Publish at: 2025-08-01

Load forecasting of electrical parameters: an effective approach towards optimization of electric load

10.11591/ijict.v14i2.pp708-716
Debani Prasad Mishra , Rudranarayan Pradhan , Ananya Priyadarshini , Subha Ranjan Das , Surender Reddy Salkuti
The increasing need for energy and the increasing cost of electricity have prompted the development of smart energy optimization systems that can help consumers reduce their electricity consumption and minimize costs. These systems are developed on the concept of a “smart grid” which is a digitalized and intelligent energy network that provides help in the efficient distribution of energy. Load forecasting plays a crucial role in the precise prediction of uncontrollable electrical load. Long-term load analysis predicts a load of more than one year and helps in the planning of power systems whereas short-term and medium-term load forecasting helps in the supply and distribution of load, maintenance of load system, ensuring safety, continuous electricity generation, and cost management. Machine learning (ML) focuses on the development of smart energy optimization systems by enabling intuitive decision-making and reciprocation to sudden variations in consumer energy demands. This study focuses on the consumption of consumer electricity and provides a solution regarding the optimized methods that will predict future consumption based on previous data and help in reducing costs and preserving renewable energy. This research promotes sustainable energy usage. The use of ML models enables intelligent decision-making and accurate predictions, making the system an effective tool for managing electricity consumption.
Volume: 14
Issue: 2
Page: 708-716
Publish at: 2025-08-01

IndoBART optimization for question answer generation system with longformer attention

10.11591/ijict.v14i2.pp478-487
Peter Andrew , Abba Suganda Girsang
The Incorporation of Question Answering system holds immense potential for addressing Indonesia’s educational disparities between the abundance of high school students and the limited number of teachers in Indonesia. These studies aim to enhance the Question Answering System model tailored for the Indonesian language dataset through enhancements to the Indonesian IndoBART model. Improvement was done by incorporating Longformer’s sliding windows attention mechanism into the IndoBART model, it would increase model proficiency in managing extended sequence tasks such as question answering. The dataset used in this research was TyDiQA multilingual dataset and translated the SQuADv2 dataset. The evaluation indicates that the Longformer-IndoBART model outperforms its predecessor on the TyDiQA dataset, showcasing an average 26% enhancement across F1, Exact Match, BLEU, and ROUGE metrics. Nevertheless, it experienced a minor setback on the SQuAD v2 dataset, leading to an average decrease of 0.6% across all metrics.
Volume: 14
Issue: 2
Page: 478-487
Publish at: 2025-08-01

Comparing bidirectional encoder representations from transformers and sentence-BERT for automated resume screening

10.11591/ijai.v14.i4.pp3404-3411
Asmita Deshmukh , Anjali Raut Dahake
In today’s digital age, organizations face the daunting challenge of efficiently screening an overwhelming number of resumes for job openings. This study investigates the potential of two state-of-the-art natural language processing models, bidirectional encoder representations from transformers (BERT) and sentence-BERT (S-BERT), to automate and optimize the resume screening process. The research addresses the need for accurate, efficient, and unbiased candidate evaluation by leveraging the power of these transformer-based language models. A comprehensive comparison between BERT and S-BERT is performed, evaluating their performance across multiple metrics, including accuracy, screening time, correlation with job descriptions, and ranking quality. The findings reveal that S-BERT outperforms BERT, achieving higher accuracy (90% vs. 86%), faster screening time (0.061 seconds vs. 1 second per resume), and stronger correlation with job descriptions (0.383855 vs. 0.1249). S-BERT though has a smaller vector size of 384 enables capturing richer semantic information compared to BERT’s vector size of 768, contributing to its superior performance. The study provides insights into the strengths and limitations of each model, offering valuable guidance for organizations seeking to streamline their talent acquisition processes and enhance candidate selection through automated systems.
Volume: 14
Issue: 4
Page: 3404-3411
Publish at: 2025-08-01

A survey of missing data imputation techniques: statistical methods, machine learning models, and GAN-based approaches

10.11591/ijai.v14.i4.pp2876-2888
Rifaa Sadegh , Ahmed Mohameden , Mohamed Lemine Salihi , Mohamedade Farouk Nanne
Efficiently addressing missing data is critical in data analysis across diverse domains. This study evaluates traditional statistical, machine learning, and generative adversarial network (GAN)-based imputation methods, emphasizing their strengths, limitations, and applicability to different data types and missing data mechanisms (missing completely at random (MCAR), missing at random (MAR), missing not at random (MNAR)). GAN-based models, including generative adversarial imputation network (GAIN), view imputation generative adversarial network (VIGAN), and SolarGAN, are highlighted for their adaptability and effectiveness in handling complex datasets, such as images and time series. Despite challenges like computational demands, GANs outperform conventional methods in capturing non-linear dependencies. Future work includes optimizing GAN architectures for broader data types and exploring hybrid models to enhance imputation accuracy and scalability in real-world applications.
Volume: 14
Issue: 4
Page: 2876-2888
Publish at: 2025-08-01

Comparative evaluation of left ventricle segmentation using improved pyramid scene parsing network in echocardiography

10.11591/ijai.v14.i4.pp3214-3227
Jin Wang , Sharifah Aliman , Shafaf Ibrahim
Automatic segmentation of the left ventricle is a challenging task due to the presence of artifacts and speckle noise in echocardiography. This paper studies the ability of a fully supervised network based on pyramid scene parsing network (PSPNet) to implement echocardiographic left ventricular segmentation. First, the lightweight MobileNetv2 was selected to replace ResNet to adjust the coding structure of the neural network, reduce the computational complexity, and integrate the pyramid scene analysis module to construct the PSPNet; secondly, introduce dilated convolution and feature fusion to propose an improved PSPNet model, and study the impact of pre-training and transfer learning on model segmentation performance; finally, the public data set challenge on endocardial three-dimensional ultrasound segmentation (CETUS) was used to train and test different backbone and initialized PSPNet models. The results demonstrate that the improved PSPNet model has strong segmentation advantages in terms of accuracy and running speed. Compared with the two classic algorithms VGG and Unet, the dice similarity coefficient (DSC) index is increased by an average of 7.6%, Hausdorff distance (HD) is reduced by 2.9%, and the mean intersection over union (mIoU) is improved by 8.8%. Additionally, the running time is greatly shortened, indicating good clinical application potential.
Volume: 14
Issue: 4
Page: 3214-3227
Publish at: 2025-08-01

Leveraging machine learning for column generation in the dial-a-ride problem with driver preferences

10.11591/ijai.v14.i4.pp2826-2838
Sana Ouasaid , Mohammed Saddoune
The dial-a-ride problem (DARP) is a significant challenge in door-to-door transportation, requiring the development of feasible schedules for transportation requests while respecting various constraints. This paper addresses a variant of DARP with time windows and drivers’ preferences (DARPDP). We introduce a solution methodology integrating machine learning (ML) into a column generation (CG) algorithm framework. The problem is reformulated into a master problem and a pricing subproblem. Initially, a clustering-based approach generates the initial columns, followed by a customized ML-based heuristic to solve each pricing subproblem. Experimental results demonstrate the efficiency of our approach: it reduces the number of the new generated columns by up to 25%, accelerating the convergence of the CG algorithm. Furthermore, it achieves a solution cost gap of only 1.08% compared to the best-known solution for large instances, while significantly reducing computation time.
Volume: 14
Issue: 4
Page: 2826-2838
Publish at: 2025-08-01

Automatic identification of native trees using MobileNetV2 model

10.11591/ijict.v14i2.pp416-426
Melidiossa V. Pagudpud , Reynold A. Rustia , Wilyn S. Marzo , Joel G. Carig
In protecting our biodiversity, knowledge of tree species is vital. However, not all people are familiar with the trees present in the community which can affect their ability to fully protect the trees. In this premise that the researchers decided to conduct this study to support the sustainable forest management project in the Province of Quirino through the creation of a model of automatic identification of native trees, using the leaves of the trees, found within the Quirino Forest landscape. The model aims to help residents with accessible tools for tree identification which can be used in the conservation efforts within the province. Transfer learning for deep learning, one of the latest advancements in image processing, shows potential for tree identification because the method dodges the labor intensive feature engineering. Using the Quirino Province native trees leaf/leaflet images dataset, which was annotated by foresters, the MobileNetV2 convolutional neural network was evaluated systemically in this paper. The result shows that the best model version to classify the native trees based on their leaves or leaflets is the one produced using 800 training steps which yields an overall accuracy of 89.61%. The result attained for the tree identification indicates that the proposed technique might be an appropriate tool to assist humans in the identification of native trees found within the landscape of Quirino and can provide reliable technical support for sustainable forest management.
Volume: 14
Issue: 2
Page: 416-426
Publish at: 2025-08-01

Enhancing database query interpretation: a comparative analysis of semantic parsing models

10.11591/ijict.v14i2.pp467-477
Gunjan Keswani , Manoj B. Chandak
The rapid proliferation of NoSQL databases in various domains necessitates effective parsing models for interpreting NoSQL queries, a fundamental aspect often overlooked in database management research. This paper addresses the critical need for a comprehensive understanding of existing semantic parsing models tailored for NoSQL query interpretation. We identify inherent issues in current models, such as limitations in precision, accuracy, and scalability, alongside challenges in deployment complexity and processing delays. This review is pivotal, shedding light on the intricacies and inefficiencies of existing systems, thereby guiding future advancements in NoSQL database querying. This methodical comparison of these models across key performance metrics-precision, accuracy, recall, delay, deployment complexity, and scalability-reveals significant disparities and areas for improvement. By evaluating these models against both individual and combined parameters, we identify the most effective methods currently available. The impact of this work is far-reaching, providing a foundational framework for developing more robust, efficient, and scalable parsing models. This, in turn, has the potential to revolutionize the way NoSQL databases are queried and managed, offering significant improvements in data retrieval and analysis. Through this paper, we aim to bridge the gap between theoretical model development and practical database management, paving the way for enhanced data processing capabilities in diverse NoSQL database applications.
Volume: 14
Issue: 2
Page: 467-477
Publish at: 2025-08-01

Techniques of deep learning neural network-based building feature extraction from remote sensing images: a survey

10.11591/ijict.v14i2.pp614-624
Shrinivas B. Khandare , Manoj B. Chandak
Recently, due to earthquake disaster, many people have lost their lives and homes, and not able to settle to new locations immediately. Therefore, a framework or a plan should be ready to immediately relocate the people to different locations or do resettlement. Much research has been done in this field but still there are problems of identifying clear building boundaries, rectangular houses, due to the problem of different shapes of the buildings. These techniques were explored for identification of clear building boundaries, rectangular houses, buildings which are more highlighted and smaller size buildings for pre-disaster and post-disaster building extraction scenarios. In this survey of building extraction techniques, most of the approach is training the network, second approach is refining the trained output features, running the trained samples on the predefined models of neural network. Several issues and their assessment are studied in these techniques. These are beneficial to the various researchers for different building extractions.
Volume: 14
Issue: 2
Page: 614-624
Publish at: 2025-08-01

Consumer behavior switching from human agents to chatbots in the health service industry

10.11591/ijict.v14i2.pp355-365
Dwi Fajar Yulianto , Titik Pratiwi , Fatih Akbarul Irsan , Faranita Mustikasari
Artificial intelligence (AI) technology is used in organizations to replace human services with technology, altering customer service experiences. Only a limited number of studies have explored how consumers change their behavior from human-assisted to technology-assisted services when using AI in frontline and specialty healthcare services. This study examined the elements that impact consumers’ transition from human agents to AI-based conversational agents using the push-pull mooring framework. Data from 147 healthcare users was evaluated using structural equation modeling. The data indicates that push effects, specifically adaptability, have a negative impact on switching behavior, while pull effects, such as responsiveness and accessibility, have a positive impact on the switching behavior of customers.
Volume: 14
Issue: 2
Page: 355-365
Publish at: 2025-08-01

Improving firewall performance using hybrid of optimization algorithms and decision trees classifier

10.11591/ijai.v14.i4.pp2839-2848
Mosleh M. Abualhaj , Ahmad Adel Abu-Shareha , Sumaya Nabil Al-Khatib , Adeeb M. Alsaaidah , Mohammed Anbar
One of the primary concerns of governments, corporations, and even individual users is their level of online protection. This is because a large number of attacks target their primary assets. A firewall is a critical tool that almost every organization uses to protect its assets. However, firewalls become less reliable when they deal with large amounts of data. One method for reducing the amount of data and enhancing firewall performance is feature selection. The main aim of this study is to enhance the firewall's performance by proposing a new feature selection method. The proposed feature selection method combines the strengths of Harris Hawks optimization (HHO) and whale optimization algorithm (WOA). Experiments were performed utilizing the NSL-KDD dataset to measure the effectiveness of the proposed method. The experiments employed the decision trees (DTs) as a machine classifier. The experimental results show that the achieved accuracy is 98.46% when using HHO/WOA for feature selection and DT for classification, outperforming the HHO and WOA when used separately for feature selection. The study's findings offer insightful information for researchers and practitioners looking to improve firewall effectiveness and efficiency in defending internet connections against changing threats.
Volume: 14
Issue: 4
Page: 2839-2848
Publish at: 2025-08-01

Contract-based federated learning framework for intrusion detection system in internet of things networks

10.11591/ijai.v14.i4.pp3324-3333
Yuris Mulya Saputra , Divi Galih Prasetyo Putri , Jimmy Trio Putra , Budi Bayu Murti , Wahyono Wahyono
A plethora of national vital infrastructures connected to internet of things (IoT) networks may trigger serious data security vulnerabilities. To address the issue, intrusion detection systems (IDS) were investigated where the behavior and traffic of IoT networks are monitored to determine whether malicious attacks or not occur through centralized learning on a cloud. Nonetheless, such a method requires IoT devices to transmit their local network traffic data to the cloud, thereby leading to data breaches. This paper proposes a federated learning (FL)-based IDS on IoT networks aiming at improving the intrusion detection accuracy without privacy leakage from the IoT devices. Specifically, an IoT service provider can first motivate IoT devices to participate in the FL process via a contract-based incentive mechanism according to their local data. Then, the FL process is executed to predict IoT network traffic types without sending IoT devices’ local data to the cloud. Here, each IoT device performs the learning process locally and only sends the trained model to the cloud for the model update. The proposed FL-based system achieves a higher utility (up to 44%) than that of a non-contract-based incentive mechanism and a higher prediction accuracy (up to 3%) than that of the local learning method using a real-world IoT network traffic dataset.
Volume: 14
Issue: 4
Page: 3324-3333
Publish at: 2025-08-01
1 2 3
...
2097 2098
Show 261 of 2098

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