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

30,411 Article Results

Two-steps feature selection for detection variant distributed denial of services attack in cloud environment

10.11591/ijai.v14.i5.pp3945-3957
Kurniabudi Kurniabudi , Eko Arip Winanto , Sharipuddin Sharipuddin
The prevalence of cloud computing among organizations poses a significant problem in ensuring security. Specifically, distributed denial of services (DDoS) attacks targeting cloud computing networks can lead to financial losses for consumers of cloud computing services. This assault has the potential to render cloud services inaccessible. The detection system serves as a remedy to prevent more substantial losses. This research aims to enhance the efficacy of the system detection model by integrating feature selection with three machine learning algorithms: decision tree (DT), random forest (RF), and naïve Bayes (NB). Therefore, our study suggests combining two phases of feature selection into the DDoS attack detection procedure. The first phase uses the information gain (IG) feature selection technique approach, and the second phase uses the principal component analysis (PCA) feature extraction approach. The technique is referred to as two-step feature selection. The test findings indicate that the implementation of two-step feature selection can enhance the performance of the DT and RF detection models by around 9%.
Volume: 14
Issue: 5
Page: 3945-3957
Publish at: 2025-10-01

Recommendation system for football player recruitment using k-nearest neighbor

10.11591/ijai.v14.i5.pp3847-3857
Maukar Maukar , Rodiah Rodiah
In modern professional football, achieving a competitive edge depends not only on on-field performance but also on effective off-field strategies, particularly in player recruitment. This study proposes a machine learning-based recommendation system to support talent identification and optimal player placement using statistical performance data. The model analyzes a wide range of features, including shots, expected goals, expected assists, pass types, offensive contributions, and defensive actions across field zones. The dataset undergoes preprocessing steps such as normalization (per 90 minutes) and dimensionality reduction. A key innovation of this research is the use of principal component analysis (PCA) to reduce feature dimensionality, minimizing redundancy while retaining essential information, which improves model efficiency and scalability. The refined data is then processed using the k-nearest neighbors (KNN) algorithm with cosine similarity, allowing the system to identify players with similar performance profiles based on directional similarity in a high-dimensional space. This combination enhances recommendation accuracy by focusing on performance structure rather than raw values. The resulting system provides actionable insights into player suitability and potential, offering clubs a data-driven tool for informed scouting and recruitment decisions. The approach demonstrates the effectiveness of combining PCA and KNN in optimizing football player recommendation systems.
Volume: 14
Issue: 5
Page: 3847-3857
Publish at: 2025-10-01

Optimized ensemble framework for predicting hydroponic stock and sales using machine learning

10.11591/ijai.v14.i5.pp3879-3886
Viktor Handrianus Pranatawijaya , Ressa Priskila , Putu Bagus Adidyana Anugrah Putra , Nova Noor Kamala Sari , Efrans Christian , Septian Geges , Novera Kristianti
The increasing global demand for food necessitates the adoption of sustainable agricultural practices. Hydroponic farming, while efficient in resource utilization, faces challenges in accurately predicting stock levels and sales due to dynamic, ever-changing factors. This research presents an optimized ensemble framework for forecasting hydroponic stock levels and sales by integrating linear regression (LR), random forest (RF), and XGBoost, further enhanced through an evolutionary algorithm (EA). The proposed framework is evaluated using root mean square error (RMSE) and mean absolute error (MAE), demonstrating significant accuracy improvements over individual models. The ensemble model achieves an RMSE reduction of 43.82% for stock prediction and 55.3% for sales forecasting compared to the best-performing individual model. Additionally, local interpretable model-agnostic explanations (LIME) are employed to offer stakeholders clear insights into decision-making processes, such as identifying "number of harvested crops" and "sales data" as key drivers of prediction outcomes. This framework supports sustainable development goals (SDGs) 9.3, 12.3, and 12.C by promoting resource efficiency, reducing food waste, and improving small-scale farmer market access. Future research will explore real-time data integration for dynamic adaptation and further model enhancements.
Volume: 14
Issue: 5
Page: 3879-3886
Publish at: 2025-10-01

Ensemble reverse knowledge distillation: training robust model using weak models

10.11591/ijai.v14.i5.pp4162-4170
Christopher Gavra Reswara , Tjeng Wawan Cenggoro
To ensure that artificial intelligence (AI) can be aligned with humans, AI models need to be developed and supervised by humans. Unfortunately, it is possible for an AI to exceed human capabilities, which is commonly referred to as superalignment models. Thus, it raised the question of whether humans can still supervise a superalignment model, which is encapsulated in a concept called weak-to-strong generalization. To address this issue, we introduce ensemble reverse knowledge distillation (ERKD), which leverages two weaker models to supervise a more robust model. This technique is a potential solution for humans to manage a super-alignment of models. ERKD enables a more robust model to achieve optimal performance with the assistance of two weaker models. We tried to train a more robust EfficientNet model with weaker convolutional neural network (CNN) models in a supervised fashion. With this method, the EfficientNet model performed better than the model trained with the standard transfer learning (STL) method. It also performed better than a model that was supervised by a single weaker model. Finally, ERKD-trained EfficientNet models can perform better than EfficientNet models that are one or even two levels stronger.
Volume: 14
Issue: 5
Page: 4162-4170
Publish at: 2025-10-01

Optimizing nitik batik classification through comparative analysis of image augmentation

10.11591/ijai.v14.i5.pp3970-3981
Suprapto Suprapto , Meilany Nonsi Tentua , Ahmad Rizki Maulana
Nitik batik is one of the most intricate and culturally significant motifs in Yogyakarta's batik tradition, characterized by its complex, geometric dot-based patterns. The unique challenges of automatically classifying nitik batik motifs stem from the high variability within the class and the limited availability of training data. This study investigates how different image data augmentation techniques can enhance the performance of a random forest classifier for nitik batik motifs. Techniques such as geometric transformations (flip, rotate, and scaling), intensity transformations (cut-out, grid mask, and random erasing), non-instance level augmentation (pairing samples), and unconditional image generation (deep convolutional generative adversarial network (DCGAN)) were used to expand the dataset and improve the model's ability to generalize. The results show that specific techniques, notably flip, cut-out, and DCGAN, significantly improved classification accuracy, with flip achieving the highest accuracy improvement of 20.20%, followed by cut-out at 19.27% and DCGAN at 16.25%. Moreover, DCGAN demonstrated the lowest standard deviation (0.78%), indicating high stability and robustness in classification performance across multiple validation folds. These findings suggest that augmentation techniques effectively improve classification accuracy and enhance the model's ability to generalize from limited and complex datasets.
Volume: 14
Issue: 5
Page: 3970-3981
Publish at: 2025-10-01

The effectiveness of ChatGPT in extracting architectural patterns and tactics

10.11591/ijai.v14.i5.pp4363-4370
Hind Milhem , Naderah Al-Jawabrah , Raghad Abu Wadi
This work investigates the potential of ChatGPT, a cutting-edge large language model (LLM), for software design analysis specifically in detecting architectural patterns and tactics. The evaluation involves comparing ChatGPT’s performance with that of Archie, a traditional Eclipse plugin designed for architectural analysis. The study uses the source code of five open-source software systems as the testing ground. Results reveal that ChatGPT achieves noteworthy performance in both pattern and tactic detection tasks. Specifically, for pattern detection, ChatGPT demonstrates an accuracy of up to 47.06%, while for tactic detection, it achieves a precision of 28.25%. While ChatGPT’s current capabilities are not yet a replacement for specialized tools like Archie, it offers significant potential as a complementary tool in architectural analysis workflows. By bridging the gap between natural language understanding and software engineering, ChatGPT could pave the way for more intelligent and automated solutions in the field. However, a key limitation is its difficulties in handling foundational or traditional tactics, resulting in a lower detection rate in certain areas. This research contributes valuable insights into the application of LLMs in software engineering, highlighting both the strengths and the limitations of ChatGPT in addressing complex architectural tasks.
Volume: 14
Issue: 5
Page: 4363-4370
Publish at: 2025-10-01

BonoNet: a deep convolutional neural network for recognizing bangla compound characters

10.11591/ijai.v14.i5.pp4171-4180
Kazi Rifat Ahmed , Nusrat Jahan , Adiba Masud , Nusrat Tasnim , Sazia Sharmin , Nusrat Jahan Mim , Imran Mahmud
The bangla alphabet includes vowels, consonants, and compound symbols. The compound nature of bangla is a product of combining two or more root bangla characters into one graph. They are difficult to differentiate because they have a sophisticated geometric shape and an immense variety of scripts used by different places and individuals. This is one of the greatest challenges in creating effective optical character recognition (OCR) systems for bangla. In this paper, a deep convolutional neural network (DCNN)-based system is presented to identify bangla compound characters with high precision. The model was trained using the AIBangla dataset. It has about 171 classes of bangla compound characters. A DCNN system, BonoNet, was designed to classify compound characters. BonoNet outperformed all other state-of-the-art architecture on the test set and improved over current state-of-the-art architecture methods. BonoNet will greatly improve the automation and analysis of the bangla language by accurately identifying these compound complex characters.
Volume: 14
Issue: 5
Page: 4171-4180
Publish at: 2025-10-01

Backpropagation neural networks for solving gas flow equations in porous media

10.11591/ijai.v14.i5.pp3744-3756
Adrianto Adrianto , Zuher Syihab , Sutopo Sutopo , Taufan Marhaendrajana
This study proposes a backpropagation neural network (BPNN) as an alternative solver for nonlinear equations in gas flow simulation through porous media. Conventional solvers like the Newton-Raphson (N-R) method are accurate but may become inefficient for large-scale or heterogeneous systems. We develop a feedforward BPNN architecture with adaptive learning rates to solve discretized residual equations from the one-dimensional gas flow model. The methodology includes finite difference discretization and mapping the nonlinear algebraic system into a four-layer neural network. The BPNN solver is validated against the Newton method across various grid sizes and heterogeneous permeability-porosity distributions. Results show that BPNN achieves high accuracy, with maximum absolute errors (MAE) of only 0.241 psi in the homogeneous model and 0.0418 psi in the heterogeneous model. While the BPNN requires more iterations and longer computation time, especially for finer grids, it exhibits the ability to learn pressure patterns and improve efficiency over time. This approach demonstrates that BPNN can serve as a viable nonlinear solver in reservoir simulation, offering flexibility in handling nonlinearities while maintaining accuracy.
Volume: 14
Issue: 5
Page: 3744-3756
Publish at: 2025-10-01

Learning assistance module based on a small language model

10.11591/ijai.v14.i5.pp4202-4210
Marco Antonio Jinete , Robinson Jiménez-Moreno , Anny Astrid Espitia-Cubillos
This paper presents the development of a low-cost learning assistant embedded in an NVIDIA Jetson Xavier board that uses speech and gesture recognition, together with a long language model for offline work. Using the large language model (LLM) Phi-3 Mini (3.8B) model and the Whisper (model base) model for automatic speech recognition, a learning assistant is obtained under a compact and efficient design based on extensive language model architectures that give a general answer set of a topic. Average processing times of 0.108 seconds per character, a speech transcription efficiency of 94.75%, an average accuracy of 9.5/10 and 8.5/10 in the consistency of the responses generated by the learning assistant, a full recognition of the hand raising gesture when done for at least 2 seconds, even without fully extending the fingers, were obtained. The prototype is based on the design of a graphical interface capable of responding to voice commands and generating dynamic interactions in response to the user's gesture detection, representing a significant advance towards the creation of comprehensive and accessible human-machine interface solutions.
Volume: 14
Issue: 5
Page: 4202-4210
Publish at: 2025-10-01

Detection of chronic kidney disease based on ensemble approach with optimal feature selection using machine learning

10.11591/ijai.v14.i5.pp4017-4031
Deepika Amol Ajalkar , Jyoti Yogesh Deshmukh , Mayura Vishal Shelke , Shalini Vaibhav Wankhade , Shwetal Kishor Patil
Chronic kidney disease (CKD) poses a significant health risk globally, necessitating early and accurate detection to ensure timely intervention and effective treatment. This study presents an advanced ensemble machine learning (ML) approach combined with optimal feature selection to enhance the detection of CKD. Using five baseline ML classifiers like gradient boosting (GB), random forest (RF), K-nearest neighbors (KNN), support vector machine (SVM), and decision tree (DT), and utilizing grid search for hyperparameter tuning, the proposed ensemble model capitalizes on the strengths of each algorithm. Our approach was tested on a public benchmark CKD dataset from Kaggle. The experimental results demonstrate that the ensemble model consistently outperforms individual classifiers and existing methods, achieving 97.5% accuracy, precision, recall, and an F1-score of 97.4%. This superior performance underscores the ensemble model's potential as a reliable early CKD detection tool. Integrating ML into CKD diagnostics enhances accuracy. It facilitates the development of automated, scalable diagnostic tools, aiding healthcare professionals in making informed decisions and ultimately improving patient outcomes.
Volume: 14
Issue: 5
Page: 4017-4031
Publish at: 2025-10-01

Design and analysis of reinforcement learning models for automated penetration testing

10.11591/ijai.v14.i5.pp4061-4073
Suresh Jaganathan , Mrithula Kesavan Latha , Krithika Dharanikota
Our paper proposes a framework to automate penetration testing by utilizing reinforcement learning (RL) capabilities. The framework aims to identify and prioritize vulnerable paths within a network by dynamically learning and adapting strategies for vulnerability assessment by acquiring the network data obtained from a comprehensive network scanner. The study evaluates three RL algorithms: deep Q-network (DQN), deep deterministic policy gradient (DDPG), and asynchronous episodic deep deterministic policy gradient (AE-DDPG) in order to compare their effectiveness for this task. DQN uses a learned model of the environment to make decisions and is hence called model-based RL, while DDPG and AE-DDPG learn directly from interactions with the network environment and are called model-free RL. By dynamically adapting its strategies, the framework can identify and focus on the most critical vulnerabilities within the network infrastructure. Our work is to check how well the RL technique picked security vulnerabilities. The identified vulnerable paths are tested using Metasploit, which also confirmed the accuracy of the RL approach's results. The tabulated findings show that RL promises to automate penetration testing tasks.
Volume: 14
Issue: 5
Page: 4061-4073
Publish at: 2025-10-01

Residual edge dense enhanced module network: a deep learning approach with multi-class SVM for lung tumor stage classification

10.11591/ijai.v14.i5.pp4032-4042
Prabakaran Jayaraman , Pandiaraj Selvaraj , Ashwini Elango
Lung cancer segmentation with positron emission tomography (PET) and computed tomography (CT) images plays a critical role to accurately detect lung cancer. Nevertheless, lung tumor segmentation in PET/CT images were extremely difficult due to the movement caused by respiration. Despite this fact, the lung tumor images shown large number of variations mostly in PET images and CT images. As PET-CT images are acquired concurrently the shape and size of lung tumor varies according to modality. To address these issues, we developed a residual edge dense enhanced module network (REDEM-NET) framework for lung tumor stage classification. The proposed REDEM-NET can process PET and CT images as inputs. In addition, the dense residual convolutional network (DRCN) collects both inputs and extracts high-dimensional features concurrently. The extracted features from both imaging modalities were fed into UNet+++ to obtain multi-level decoded features. The extracted decoded features are concurrently supplied to the pixel level learning module (PELM) and edge level learning module (E2LM) which resulting in two outputs for subsequent learning. The outputs were merged to provide a very precise lung tumor segmentation. Furthermore, segmented tumor was fed to multi-class support vector machine (MC-SVM) for lung tumor stage classification. Moreover, it was able to identify three stages and its substages namely primary tumor, region lymph node and distant metastasis.
Volume: 14
Issue: 5
Page: 4032-4042
Publish at: 2025-10-01

Intent detection in AI chatbots: a comprehensive review of techniques and the role of external knowledge

10.11591/ijai.v14.i5.pp4250-4259
Jemimah K. , Rajkumar Kannan , Frederic Andres
Artificial intelligence (AI) chatbots have become essential across various industries, including customer service, healthcare, education, and entertainment, enabling seamless, and intelligent user interactions. A key component of chatbot functionality is intent detection, which determines the underlying purpose of user queries to provide relevant responses. Traditional intent detection methods, such as rule-based and statistical approaches, often struggle with adaptability, especially in complex, dynamic conversations. This review examines the evolution of intent detection techniques, from early methods to modern deep learning and knowledge-enriched models. It introduces the domain type-conversation turns-adaptivity-external knowledge (DCAD) classification, highlighting its significance in improving chatbot accuracy and contextual awareness. The paper categorizes existing intent detection models, analyzes their applications across various sectors, and discusses key challenges, including data integration, language ambiguity, and ethical concerns. By exploring emerging trends and future directions, this review underscores the critical role of external knowledge in enhancing chatbot performance and user experience.
Volume: 14
Issue: 5
Page: 4250-4259
Publish at: 2025-10-01

Multiclass instance segmentation optimization for fetal heart image object interpretation

10.11591/ijai.v14.i5.pp4137-4150
Hadi Syaputra , Siti Nurmaini , Radiyati Umi Partan , Muhammad Taufik Roseno
This research aims to develop a multi-class instance segmentation model for segmenting, detecting, and classifying objects in fetal heart ultrasound images derived from fetal heart ultrasound videos. Previous studies have performed object detection on fetal heart images, identifying nine anatomical classes. Further, these studies have conducted instance segmentation on fetal heart images for six anatomical classes. This research seeks to expand the scope by increasing the number of classes to ten, encompassing four main chambers left atrium (LA), right atrium (RA), left ventricle (LV), right ventricle (RV); four valves tricuspid valve (TV), pulmonary valve (PV), mitral valve (MV), and aortic valve (AV); one aorta (Ao), and the spine. By developing an instance segmentation method for segmenting ten anatomical structures of the fetal heart, this research aims to make a significant contribution to improving medical image analysis in healthcare. It also aims to pave the way for further research on fetal heart diseases using AI. The instance segmentation approach is expected to enhance the accuracy of segmenting fetal heart images and allow for more efficient identification and labeling of each anatomical structure in the fetal heart.
Volume: 14
Issue: 5
Page: 4137-4150
Publish at: 2025-10-01

Comparison of HSV-color and ANN-HSV-color segmentation for detecting soybean adulteration

10.11591/ijai.v14.i5.pp3734-3743
Farid Rahmat Abadi , Rudiati Evi Masithoh , Lilik Sutiarso , Sri Rahayoe
Soybeans are an important food crop, but their quality is often compromised by contamination with other materials, a process known as adulteration. Conventional methods for detecting adulteration are slow; therefore, there is a need for rapid and non-invasive alternatives. This study aimed to assess the capability of hue-saturation-value (HSV) color segmentation and its combination with artificial neural networks (ANN) to identify adulteration in soybean samples. This research employed image processing and machine learning to segment soybeans mixed with adulterants at concentrations of 5%, 10%, 15%, 20%, and 25%. The HSV method successfully distinguished soybeans and other materials, but some challenges were observed in shadow regions and areas with similar colors. The HSV-ANN model with six hidden layers performed well with a calibration accuracy of R² value of 0.97 and root-mean-square error (RMSE) of 2.16%, which provided more detailed segmentation, although it still had some problems in shadow regions and undetected corn embryo parts. The validation results indicated that the HSV model had an R² value of 0.98 and RMSE of 4.48%, while the HSV-ANN model had an R² value of 0.96 and RMSE of 1.3%. Both models were capable of predicting the levels of adulteration, and the HSV-ANN model proved to be more accurate. It is concluded that both methods are efficient; however, there is a need for more work on modeling and sampling to increase the segmentation precision and decrease the biases, especially in the shadow and overlapped color.
Volume: 14
Issue: 5
Page: 3734-3743
Publish at: 2025-10-01
1 2 3
...
2027 2028
Show 137 of 2028

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