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

30,938 Article Results

Change detection and classification of satellite images using convolutional neural network

10.11591/ijai.v15.i1.pp329-337
Raghavendra Srinivasaiah , Santosh Kumar Jankatti , Manjunath Ramanna Lamani , Niranjana Shravanabelagola Jinachandra
Satellite and airborne imagery, collectively known as earth observation imagery, are images of the earth collected from spaceborne or airborne platforms such as satellites and aircraft. Over the last 100 years, with the fast development of aviation, space exploration, and imaging technologies, the coming together of these technologies has been inevitable. Earth observation imagery has many applications in regional planning, geology, reconnaissance, fishing, meteorology, oceanography, agriculture, biodiversity conservation, forestry, landscape, intelligence, cartography, education, and warfare. With the rise in the number of these airborne and spaceborne imaging platforms being deployed by government and private entities alike, the capability to sift through and analyze vast amounts of data generated by these platforms is the need of the hour. With the exponential improvement in the computational capabilities of computers over the last half a century, analysts are exceedingly moving towards the practice of artificial intelligence, machine learning (ML), and computer vision solutions to automate a large part of the processes employed in analyzing earth observation imagery. This work recommends a workflow to perceive and classify changes in earth observation imagery of a given area by utilizing the vast flexibility that convolutional neural networks (CNN) provide.
Volume: 15
Issue: 1
Page: 329-337
Publish at: 2026-02-01

Enhancing medical language models with big data technologies

10.11591/ijai.v15.i1.pp289-299
Ayoub Allali , Ibtihal Abouchabaka , Najat Rafalia
In this study, we present an end-to-end, big-data–driven framework for continuously enriching and fine-tuning large language models (LLMs) with the latest professional and scientific medical knowledge. Streaming updates from premier sources such as The New England Journal of Medicine (NEJM) are ingested via an Apache Kafka cluster for low-latency delivery and durably archived in a three-node Apache Hadoop (Hadoop distributed file system (HDFS)) system. Each new article is preprocessed into high dimensional embeddings and indexed in a Milvus vector database to enable sub-second semantic retrieval over millions of records. At query or batch time, our retrieval-augmented generation (RAG) module retrieves the top-k relevant embeddings from Milvus and injects them into prompts for DeepSeek-R1, GPT-4o-mini, and Llama 3, models which are hosted, fine tuned, and served via Ollama on an NVIDIA GeForce RTX 3050 Ti GPU for efficient inference and continual learning. The enriched outputs are seamlessly delivered to end users through a Telegram bot programmed in Python using the Telebot library, linking the RAG-enhanced LLMs to an intuitive chat interface. Our Kafka, HDFS, Milvus, RAG, LLM, or Telegram bot pipeline demonstrably improves factual accuracy and topical currency of AI-generated medical insights across clinical decision support, patient engagement and education, drug discovery and development, virtual health assistants, and mental health support, laying the groundwork for truly intelligent, responsive, and data-driven healthcare solutions.
Volume: 15
Issue: 1
Page: 289-299
Publish at: 2026-02-01

Real-time object detection to classify export quality of mangosteen using variants of you only look once version 8

10.11591/ijai.v15.i1.pp116-128
Dian Sa'adillah Maylawati , Mi’raj Fuadi , Kurniawan Yniarto , Yuhendra AP , Rizky Rahmat Nugraha , Akbar Hidayatullah Harahap , Agung Wahana
Mangosteen is one of the leading export commodities from Indonesia. Despite its great economic potential, only about 25% of Indonesian mangosteens meet export standards, mainly due to visual defects such as yellow sap and spots on the skin of the fruit. The process of sorting export worthy mangosteens has been done manually, which tends to be time consuming and inconsistent. Therefore, this study aims to utilize artificial intelligence technology in building a real-time image recognition model to improve the efficiency and accuracy of the export-quality mangosteen sorting process. This study uses you only look once version 8 (YOLOv8) as an image recognition model with YOLOv8 variants, including nano, small, medium, large, and extra large variants. The results of the study using 4,014 primary and 255 secondary data of mangosteen, the highest performance is reached by YOLOv8 medium 82% of accuracy, 0.856 of mean average precision (mAP)50, and 0.616 of mAP50-95. This result is obtained from 70% training, 20% validation, and 10% testing data with epoch stop 85. These results indicate that the model can provide good performance in mangosteen export quality classification. This research contributes to the fields of agricultural technology and artificial intelligence by offering an innovative solution to a practical problem, enhancing efficiency, accuracy, and scalability in export-quality mangosteen sorting.
Volume: 15
Issue: 1
Page: 116-128
Publish at: 2026-02-01

A comprehensive survey of cyberbullying on social media: challenges, detection, and AI-based prevention

10.11591/ijai.v15.i1.pp86-96
Ammar Odeh , Osama Alhaj Hassan , Anas Abu Taleb , Abobakr Aboshgifa , Nabil Belhaj
Cyberbullying is a pervasive issue in the digital landscape, particularly on social media platforms, where individuals engage in online harassment, intimidation, and abuse. Unlike traditional bullying, cyberbullying has a broader reach, anonymity, and persistence, making it a growing concern for mental health, social well-being, and online safety. This paper provides a comprehensive survey of cyberbullying trends, its psychological and social impacts, and the role of social media in amplifying the problem. It explores existing detection and prevention strategies, including artificial intelligence (AI)-driven approaches, policy frameworks, and platform-based moderation techniques. Furthermore, it discusses challenges in enforcement, the limitations of automated detection systems, and the need for improved legal measures. This paper uniquely contributes an integrated perspective on cyberbullying detection and prevention by synthesizing current research across psychological, sociocultural, and technical dimensions. It emphasizes underexplored gaps such as multilingual detection, real-time moderation, and cross-platform enforcement, and proposes a layered framework to guide future research and policy.
Volume: 15
Issue: 1
Page: 86-96
Publish at: 2026-02-01

Gradient descent optimization based weighted federated learning for privacy-preserving framework

10.11591/ijai.v15.i1.pp878-887
Gururaj Prakash Murthy , Chandrashekhar Pomu Chavan
Federated learning (FL) is a disseminated machine learning (ML) paradigm that gained significant consideration in modern days, particularly in a domain of the internet of things (IoT). FL saves communication bandwidth when compared to centralized ML processes by eliminating the need to transmit raw client data to a central server, thereby enhancing data privacy. Nevertheless, participant privacy is still compromised through inference attacks and similar threats. Additionally, a data excellence provided through clients can differs significantly, and excessive inclusion of low-quality data during training may degrade the overall performance of the global model. Hence, this research introduces a gradient descent optimization assisted weighted federated learning (GDO-WFL) method for privacy preservation. The proposed GDO-WFL approach is significantly efficient as it strengthens privacy preservation through reducing exposure to inference attacks and optimises gradient updates for secure learning. Through weighting client contributions based on data quality, an undesirable effect of low-quality data can be minimised, helping to maintain a strength as well as accuracy of the global model. The experimental results illustrate a proposed GDO-WFL approach maintains an overall accuracy of 99.3 and 91.5% on MNIST and CIFAR-10 datasets as compared to the existing method of FedlabX method.
Volume: 15
Issue: 1
Page: 878-887
Publish at: 2026-02-01

Evaluating the detected communities using traditional algorithms on keyword co-occurrence networks

10.11591/ijai.v15.i1.pp919-928
Kiruthika R. , Krishnaveni Sakkarapani
Community detection is one of the most significant research areas in network analysis, which helps to understand the internal structure of large networks. This work utilizes the traditional community detection methods on a keyword co-occurrence graph derived from the Scopus bibliographic database. This research article primarily focused on the index keywords of deep learning driven publications obtained from three major network Scopus bibliometric datasets (SBD), namely SBD_1 as 2006-2013, SBD_2 as 2014-2016, and SBD_3 as 2017. For this proposed model framework, the existing traditional algorithms, including Louvain, greedy modularity optimization (GMO), Leiden, Infomap, speaker-listener label propagation algorithm (SLPA), Walktrap, SpinGlass, K-Clique, and Clauset, Newman and Moore (CNM) methods are applied to detect communities from the network and carried out through Python. Comparisons among these algorithms, Leiden, SpinGlass, and Louvain are considered as better algorithms for our work based on the detected communities, modularity score and other metrics to evaluate the performance of detected communities from the network. This research proposes an ideology for the selection process of algorithms that depends on different factors like network characteristics, network structure, dataset size, and computational efficiency. This analysis suggests a unique perspective on the effectiveness of each method in the Scopus bibliometric network and its potential to enhance research topic exploration.
Volume: 15
Issue: 1
Page: 919-928
Publish at: 2026-02-01

A review of modern techniques for plant disease identification and weed detection in precision agriculture

10.11591/ijai.v15.i1.pp998-1008
Mohammad Naseera , Arpita Gupta
Plant disease identification and weed detection are critical components of precision agriculture, aimed at ensuring high crop yields and sustainable farming practices. These processes involve the use of advanced machine learning and deep learning techniques to automatically identify and classify plant diseases and distinguish between crops and weeds in agricultural fields. Traditional methods for managing these challenges are often labor intensive, prone to errors, and environmentally unsustainable, necessitating the development of automated, accurate, and scalable solutions. This survey provides a comprehensive review of the state-of-the-art approaches, including pixel-based, region-based, and spectral-based methods, and evaluates their effectiveness in various agricultural contexts. Additionally, it identifies significant challenges such as data scarcity, model generalization, and computational constraints, while proposing potential research directions to address these gaps. The findings aim to guide future research in developing more robust and interpretable models that can be deployed in real-world agricultural environments, ultimately contributing to more efficient, precise, and sustainable farming practices.
Volume: 15
Issue: 1
Page: 998-1008
Publish at: 2026-02-01

Detection and forecasting of mental health disorders using machine learning models on social media data

10.11591/ijai.v15.i1.pp672-680
Chaithra Indavara Venkateshagowda , Roopashree Hejjajji Ranganathasharma , Yogeesh Ambalagere Chandrashekaraiah , Narve Lakshminarayan Taranath
The detection and classification of depression and other mental disorders have become crucial in the modern era, particularly with the growing reliance on social media for self-expression. Existing systems often face challenges like limited prediction accuracy, difficulty forecasting future mental illnesses, and handling both clinical and non-clinical data. This study proposes a novel analytical model that not only screens individuals' current mental health status from social media content but also predicts the likelihood of future mental health issues. The proposed methodology integrates classical machine learning (ML) models, ensemble learning approaches, and pretrained models for enhanced detection and forecasting accuracy. The outcome shows that pre-trained language models accomplished maximized F1-score and overall performance significantly better than conventional ML and ensemble models. The system outperforms existing methods with a significant accuracy improvement, achieving 90.9% overall accuracy, a 7.2% improvement over traditional ML classifiers, 5.8% over ensemble models, and 11.3% over language models.
Volume: 15
Issue: 1
Page: 672-680
Publish at: 2026-02-01

Deep learning-based cervical cancer detection via colposcopy images integrated into an Android mobile application

10.11591/ijai.v15.i1.pp338-349
Retno Supriyanti , Arsil Kultura Anzil , Yogi Ramadhani , Suroso Suroso , Wahyu Widanarto , Muhammad Alqaaf , Kartika Dwi Hapsari , Futiat Diana Kartika
Cervical cancer is a form of cancer that develops in the cells of the cervix, the lower part of the uterus that connects the uterus to the vagina. Early detection is essential for improving the chances of recovery from cervical cancer. One method for early detection is colposcopy image analysis, a medical procedure that examines the cervix and captures images for evaluation. These images were analyzed to observe color changes after the visual inspection with acetic acid (VIA) process. However, this analysis requires experienced and specially trained medical personnel. To address this challenge, a system that can automatically classify cervical cancer images is needed. Therefore, researchers proposed designing and developing an Android mobile application to enable early detection of cervical cancer using the convolutional neural network (CNN) algorithm. The CNN model was tested using test data to evaluate its performance. The optimized CNN model utilizing the ResNet50 architecture achieved 86% test accuracy, 85% precision, and 87% recall. The test results indicate that the model's accuracy is consistent before and after its implementation on the mobile application, confirming the effectiveness of both the model and its implementation as diagnostic tools.
Volume: 15
Issue: 1
Page: 338-349
Publish at: 2026-02-01

Enhancing vehicular ad hoc network security through a trust based vehicular model for attack mitigation

10.11591/ijai.v15.i1.pp247-256
Shilpa Shilpa , Thiruvenkadam Prasanth
In vehicular ad-hoc networks (VANETs), ensuring secure and reliable communication is essential due to the growing threat of cyber-attacks. As attacks can disrupt data transmission and compromise user privacy and network integrity, it is vital to develop robust security solutions. Hence, this work introduces a trust-based vehicular security (TVS) model, which leverages trust metrics to enhance VANET security. The main objective was to establish secure connections between vehicles and infrastructure nodes, effectively mitigating attacks while maintaining higher throughput. The methodology integrated a dynamic trust evaluation model to prevent malicious activities and ensure secure data transmission. The TVS model’s performance was compared to an existing VANET model, showing improved results in terms of detection rate, misclassification rate, and throughput. The findings demonstrate an average misclassification rate of 22.75%, a detection rate of 14.77%, and a throughput of 11.45%, highlighting the superior effectiveness of the TVS model in attack-prone environments when compared with existing VANET models. The TVS model provides a promising security solution for VANETs, offering enhanced protection against denial-of-service (DoS) attacks and spoofing (cyber-attacks) with better accuracy and network performance. The novelty lies in the dynamic, multi-trust-based approach for secure communication in vehicular networks.
Volume: 15
Issue: 1
Page: 247-256
Publish at: 2026-02-01

Interpretable artificial intelligence system for personalized cognitive stimulation

10.11591/ijai.v15.i1.pp164-176
Rubén Baena-Navarro , Yulieth Carriazo-Regino , Mario Macea-Anaya
The growing need to preserve cognitive health in aging populations has intensified interest in adaptive digital interventions that provide personalized and interpretable support. This study presents a web-based cognitive stimulation system for older adults integrating a multilayer perceptron (MLP) classifier, expert-derived symbolic rules, and explainable artificial intelligence (XAI) techniques, including Shapley additive explanations (SHAP) and local interpretable model-agnostic explanations (LIME). The platform was evaluated through a 24-week intervention involving 150 participants aged 65 years and older, combining baseline cognitive profiling, rule-guided recommendation logic, and neural prediction to support individualized task allocation. Compared with a control group, participants in the intervention arm showed statistically significant improvements in cognitive outcomes (p <0.05), with measurable gains in memory- and attention-related tasks. The explainability component enabled examination of model behavior at the level of individual features through feature attribution analysis and symbolic consistency checks, supporting interpretation beyond aggregate performance metrics. Unlike approaches dependent on high-end extended reality (XR) infrastructures or game centered interaction, the system was implemented to operate under low connectivity conditions and was tested with participants from diverse educational backgrounds. This hybrid configuration provides an interpretable basis for cognitive support initiatives adaptable to community settings contexts.
Volume: 15
Issue: 1
Page: 164-176
Publish at: 2026-02-01

Explainable rice yield from Sentinel-1 and Sentinel-2 satellite data for food security

10.11591/ijai.v15.i1.pp615-627
Dhimas Tribuana , Usman Sattar , Baharuddin Mide , Dayanti Dayanti
Reliable, explainable crop-yield estimates are essential for food-security planning in data-sparse regions. We present a transparent pipeline for district-level (regency) rice yield prediction in Indonesia that fuses Sentinel-1 synthetic aperture radar (SAR), Sentinel-2 normalized difference vegetation index (NDVI), and weather/reanalysis features. The system standardizes inputs per province, fixes a 16-day temporal key, and uses a small, auditable ensemble of tree models (gradient boosting+light gradient-boosting machine (LightGBM)). Trained on ≤2023 data and evaluated on a 2024 temporal hold-out, a joint West Java ∪ South Sulawesi model achieves root mean square error (RMSE)≈0.80 t/ha, mean absolute error (MAE)≈0.48 t/ha, and R-squared (R²)≈0.33 at regency scale. Feature importances and Shapley additive explanations (SHAP) confirm that phenology (NDVI peak, integral, green-up/senescence), SAR backscatter (vertical transmit-vertical receive/vertical transmit-horizontal receive (VV/VH)), and wind/pressure are consistent drivers under monsoon conditions. The workflow supports routine, one-click provincial updates and produces parity maps and error bars for actionable diagnostics. These results demonstrate that combining Sentinel-1, Sentinel-2, and basic meteorology delivers accurate, interpretable, and operational yield signals suited to Indonesia’s food security needs, while providing a clear recipe for scaling to additional provinces.
Volume: 15
Issue: 1
Page: 615-627
Publish at: 2026-02-01

A novel BERT-long short-term memory hybrid model for effective credit card fraud detection

10.11591/ijai.v15.i1.pp788-797
Oussama Ndama , Safae Ndama , Ismail Bensassi , El Mokhtar En-Naimi
In the rapidly evolving landscape of financial transactions, the detection of fraudulent activities remains a critical challenge for financial institutions worldwide. This study introduces a novel bidirectional encoder representation from transformers (BERT)–long short-term memory (LSTM) hybrid model that integrates both textual and numerical data to enhance credit card fraud detection. Leveraging BERT for deep contextual embeddings and LSTM for sequence analysis, the model provides a comprehensive approach that surpasses traditional fraud detection systems primarily based on numerical analysis. On the validation set, the model achieved a recall of 100% and an accuracy of 99.11%, highlighting strong effectiveness in identifying fraudulent transactions under class imbalance. Through rigorous evaluation, the model demonstrated exceptional accuracy and reliability, promising improvements in fraud detection and mitigation. This paper details the development and validation of the hybrid model, emphasizing its use of mixed data types to capture complex patterns in transaction data. The results indicate a new frontier in fraud detection by combining natural language processing (NLP) and sequential data analysis to create a robust solution for real-world applications, supporting the security and integrity of financial systems globally.
Volume: 15
Issue: 1
Page: 788-797
Publish at: 2026-02-01

Securing cloud data with machine learning: trends, gaps, and performance metrics

10.11591/ijai.v15.i1.pp44-55
Blessing Ifeoluwa Omogbehin , Tshiamo Sigwele , Thabo Semong , Aone Maenge , Zhivko Nedev , Hlomani Hlomani
The increasing reliance on cloud computing has raised significant concerns about the security of data access control, as traditional models are insufficient in managing the dynamic and large-scale nature of cloud environments. This review evaluates machine learning (ML)-based approaches to improve cloud data security, with a particular focus on advancements in anomaly detection and insider threat prevention. Deep learning (DL) models emerge as the most dominant, utilized by 47% of the studies due to their superior ability to process large datasets and adapt to real-time environments. Random forest models are also prominent, being adopted in 20% of the studies for their strong performance in anomaly detection and categorization. TensorFlow stands out as the most widely used tool, featuring in nearly 37% of the reviewed works, while datasets like Amazon Access and computer emergency response team (CERT) are employed in 20% and 13% of the research, respectively. Anomaly detection and prevention are critical priorities, accounting for 41.2% of the research objectives. However, gaps remain, with 21.7% of the studies noting adversarial vulnerabilities and 13% identifying limitations in dataset diversity. The review recommends further development of ML models to address these challenges, expanding dataset diversity, and improving real-time monitoring techniques to enhance cloud data security.
Volume: 15
Issue: 1
Page: 44-55
Publish at: 2026-02-01

Secure and interoperable electronic health record exchange using blockchain and ECDHE-based access control

10.11591/ijai.v15.i1.pp310-321
Krishna Prasad Narasimha Rao , Chinnaiyan Selvan
Electronic health records (EHRs) act as comprehensive records of health related transactions and essential resources of data in the healthcare sector. However, the integrity and security problems of EHR continue to be inflexible. The architecture of blockchain-enabled EHR addresses the issues of integrity efficiently. This paper developed the decentralized patient centric healthcare data management (PCHDM) system using blockchain enabled EHR architecture for addressing problems with access control, record privacy and data confidentiality. This framework places the patient at the control center, ensuring secure storage of EHR records and obtaining efficient data management through the integration of blockchain and interplanetary file system (IPFS). To prevent access by unauthorized users, the proposed elliptic curve Diffie-Hellman ephemeral (ECDHE) mechanism incorporates smart contract-enabled access control for managing EHR transactions and enforcing access strategies. This architecture incorporates hyperledger fabric endorsement policies (HFRP) to address scalability problems while preserving patient privacy and securing medical data. The developed method secures the EHR data and facilitates the data exchange across heterogeneous healthcare platforms, ensuring standard communication among different EHR systems. The architecture is assessed with parameters of time for block creation, the computational overhead of transaction with encryption key size and EHR upload and download time.
Volume: 15
Issue: 1
Page: 310-321
Publish at: 2026-02-01
1 2 3
...
2062 2063
Show 105 of 2063

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