Articles

An insight on using deep learning algorithm in diagnosing gastritis

10.12928/telkomnika.v23i6.27191

Ragu; Universiti Tun Hussein Onn Malaysia (UTHM) P. J. , Ashok; Universiti Tun Hussein Onn Malaysia (UTHM) Vajravelu , Muhammad Mahadi; Universiti Tun Hussein Onn Malaysia (UTHM) bin Abdul Jamil , Syed Riyaz; NITTE University Ahammed

Chronic autoimmune gastritis (CAG) is a condition in which the stomach membrane is significantly impacted by inflammation. Despite the availability of numerous modern medical techniques, the detection of this condition continues to be a difficult challenge. White light endoscopy (WLE) has been employed to diagnose gastritis, but it has been subject to certain constraints. This technique is most effective when executed by an endoscopist who possesses a high level of expertise. In the present day, WLE is frequently accompanied by artificial intelligence (AI) due to its superior ability to detect defects that lead to damage. Recently, there has been a substantial increase in the efficacy of AI in conjunction with the expertise of endoscopists in the detection of CAG. The 25,216 intriguing case studies were examined in the eight selected studies. The collection comprised 84,678 frames and 10,937 images. The AI was 94% sensitive (95% CI: 0.88-0.97, I2 = 96.2%) and 96% specific (95% CI: 0.88-0.98, I2 = 98.04%). The receiver operating characteristic curve had an area of 0.98 (95% confidence interval: 0.96–0.99). A camera is highly effective when combined with AI to assist in the identification of CAG and is advantageous for clinical review.

Volume: 23

Issue: 6

Page: 1528-1542

Publish at: 2025-12-01

Optimizing vehicle inspection efficiency and integrity in Tanzania through blockchain technology

10.12928/telkomnika.v23i6.26913

Cleverence; University of Botswana Kombe , Robert; National Institute of Transport (NIT) Sikumbili , Leticia; National Institute of Transport (NIT) Mihayo , Angela- Aida; National Institute of Transport (NIT) Runyoro

This study proposes a blockchain-based solution to improve the efficiency and integrity of vehicle inspections in Tanzania, with a focus on the National Institute of Transport. The system combines Hyperledger fabric, a permissioned blockchain that provides identity management and fine-grained access control, with the InterPlanetary file system (IPFS), a decentralized content-addressed store for large artifacts such as inspection images and portable document format (PDF) forms. Smart contracts encode inspection rules and approvals, which yield tamper-evident records, faster retrieval of histories, and uniform enforcement across centers. A mathematical model based on the M/M/1 queueing system, combined with a cost-benefit analysis, supports empirical findings: the total inspection cycle time decreases by approximately 30 percent, the average waiting time declines by about 20 to 30 percent, and annual operational savings reach approximately USD 800,000. These gains enhance auditability and transparency, which contribute to road safety outcomes by reducing opportunities for tampering and error. The design includes offline capture with later synchronization, which suits centers with intermittent connectivity. The approach is transferable to adjacent public services, for example, licensing, fine collection, and selected registries.

Volume: 23

Issue: 6

Page: 1506-1517

Publish at: 2025-12-01

Enhancing handover management in 5G networks with encoder-decoder LSTM for multistep forecasting

10.12928/telkomnika.v23i6.27107

Zineb; University of Science and Technology Mohamed Boudiaf Ziani , Mohammed; University of Science and Technology Mohamed Boudiaf Hicham Hachemi , Bouabdellah; University of Science and Technology Mohamed Boudiaf Rahmani , Mourad; University of Tlemcen Hadjila

The continuous evolution of wireless communication networks, fueled by advancements in 5G and the envisioned potential of 6G technologies, has introduced significant challenges in mobility management and handover (HO) optimization. The frequent HOs due to network densification, particularly at high frequencies like millimeter waves (mmWave) and terahertz (THz) bands, can lead to increased latency, and potential service disruptions. To address these issues, artificial intelligence (AI) driven approaches are emerging as promising alternatives. This paper explores the use of deep learning techniques for predictive HO management. An encoder-decoder long short-term memory (ED-LSTM) model is proposed to generate multistep predictions of future reference signal received power (RSRP) values. The model was trained and evaluated on two distinct real-world drive-test datasets. The results demonstrate that the proposed ED-LSTM model achieves lower prediction error, with a mean absolute error (MAE) of 2.07 for dataset 1 and 2.33 for dataset 2, and a mean absolute percentage error (MAPE) of 2.80% for dataset 1 and 2.96% for dataset 2. Overall, the ED-LSTM outperforms the bidirectional LSTM (BiLSTM) and standard LSTM (S-LSTM) model, achieving improvements of 33–38% on dataset 1 and 48-50% on dataset 2 in terms of MAE and MAPE, respectively.

Volume: 23

Issue: 6

Page: 1518-1527

Publish at: 2025-12-01

Modeling chemical kinetics of geopolymers using physics informed neural network

10.11591/ijict.v14i3.pp822-829

Blesso Abraham , Thirumalaivasal Devanathan Sudhakar

Using a physics informed neural network for the analysis of geopolymers as an alternate material for cement can be a viable approach, as neural networks are capable of modeling complex, nonlinear relationships in data, which can be beneficial for representing the dynamics of chemical properties. If you have a substantial amount of theoretical data, a neural network can learn patterns and relationships in the data, even when the underlying system dynamics are not well-defined or are difficult to model analytically. A welltrained neural network can generalize from the training data to make predictions for unseen scenarios, which can be useful for real-time analysis of the material.

Volume: 14

Issue: 3

Page: 822-829

Publish at: 2025-12-01

Particle swarm optimization-optimized integrator backstepping for the control of electric wheelchairs velocity

10.12928/telkomnika.v23i6.27516

Djamila; University of ORAN 1 Boubekeur , Khayreddine; Tlemcen University Saidi , Mohammed; Tlemcen University Messirdi , Abelmadjid; Tlemcen University Boumédiène

Most people suffering from temporary or permanent disabilities rely on wheelchairs or electric powered wheelchairs (EPW) to maintain autonomy of movement. To address different EPW control challenges, several studies have investigated this kind of robot. This paper focuses on the optimization of the integrator backstepping control parameters of the EPW. The system operates using two permanent magnet synchronous motors (PMSM), noted for their great efficiency, substantial torque, minimal noise, and robustness. At first, the dynamic model for both EPW-motors is showned. After that, a nonlinear integrator backstepping command based on Lyapunov’s second technique, which combines the choice of the energy function with the control laws, was applied to the resulting global model. To ensure optimal performance, the control parameters were tuned by means of an optimization approach. Specifically, the particle swarm optimization technique (PSO) was employed to search for the optimal parameters (gains) of the integrator backstepping controller. In order to assess the performance of the optimized backstepping–based control approach, numerical simulations were conducted to illustrate the evolution of both electrical and mechanical velocity- related variables.

Volume: 23

Issue: 6

Page: 1646-1656

Publish at: 2025-12-01

A blended ensemble approach for accurate human activity recognition

10.11591/ijai.v14.i6.pp5131-5139

Rezwana Karim , Afsana Begum , Miskatul Jannat , Abu Kowshir Bitto

Human activity recognition (HAR) is a novel computer vision area with applications in fashion, entertainment, healthcare, and urban planning. Previously, convolutional neural networks (CNNs) were used in HAR due to their ability to extract spatial features from images. However, CNNs are not effective in processing varying input sizes and long-range dependencies in complex human motions. This work examines another approach using vision transformers (ViT) and swin transformers (SwinT) that process images as patch sequences and perform self-attention. These models particularly excel in learning global relationships and minor motion changes in body motion and are therefore very well-suited to variegated and subtle activity detection. To further enhance recognition performance, we propose a hybrid ensemble method by combining ViT and SwinT models with different scales (small, base, and large). Experimental outcomes show that while single transformer models are competitive, the hybrid ensemble beats them across the board with the highest accuracy and balanced precision, recall, and F1-score. These findings confirm that the intended ensemble model provides a more scalable and robust solution than either single-model or CNN-based approaches, and this encourages accurate human activity recognition.

Volume: 14

Issue: 6

Page: 5131-5139

Publish at: 2025-12-01

Retrieval-augmented generation for Arabic legal information: the family code case study

10.12928/telkomnika.v23i6.27400

Jamal; Abdelmalek Essaâdi University Hrimech , Mohammed; Abdelmalek Essaâdi University Mghari , Youssef; Abdelmalek Essaâdi University Zaz

This document describes the implementation and evaluation of a retrieval-augmented generation (RAG) system to improve access to and understanding of Moroccan law, particularly the family code in Arabic. The research addresses the drawbacks of the widely used linguistic model applied to complex legal terminology in Arabic and aims to help citizens access crucial legal data. We built a new custom dataset with 2.5 k question-answer pairs while preprocessing and using the BGE-m3 embedding model in this experiment. Performance metrics, such as mean reciprocal rank (MRR), Recall@k, and F1-score, indicate that the RAG approach is effective compared to the use of standalone large language models (LLMs). Moreover, an evaluation on metrics such as the blue score, fidelity, response relevance, and contextual relevance indicated that the matching of meanings and context were well captured, which signifies a very good semantic understanding. The research highlights the need for language-specific model specialization in Arabic and presents its main challenges, such as dialectal variations and appropriate evaluation measures. The results indicate that well-developed RAG systems offer a promising approach to improving access to legal information in Arabic-speaking practice communities and to guiding future research and development in this field.

Volume: 23

Issue: 6

Page: 1495-1505

Publish at: 2025-12-01

Advanced signal transformation techniques to improve spectral efficiency in visible light communication systems

10.12928/telkomnika.v23i6.26835

Shahir; Al-Imam University College Fleyeh Nawaf , Ammar; Tikrit University Bouallegue , Sameh; University of Carthage Najeh

Visible light communication (VLC) offers high-speed wireless communication using the visible light spectrum. Achieving high spectral efficiency while maintaining a low bit error rate (BER) remains a challenge. This paper explores the use of quadrature amplitude modulation (QAM) combined with orthogonal frequency division multiplexing (OFDM) to address these challenges. Matrix laboratory (MATLAB) simulations show that QAM-OFDM achieves a BER of 0.001 at comparable signal-to-noise ratios (SNR), outperforming traditional hermitian symmetry (HS), complex signal mapping (CSM), and quad-light emitting diode (LED) complex modulation (QCM) techniques. Unlike CSM, and QCM, which increase complexity, and BER, QAM-OFDM efficiently utilizes available bandwidth, reducing errors, and enhancing spectral efficiency. The study concludes, that QAM-OFDM happens to be the optimal solution for the future VLC systems, offering better performance within both efficiency, and reliability.

Volume: 23

Issue: 6

Page: 1449-1456

Publish at: 2025-12-01

The bootstrap procedure for selecting the number of principal components in PCA

10.11591/ijict.v14i3.pp1136-1145

Borislava Toleva

The initial step in determining the number of principal components for both classification and regression involves evaluating how much each component contributes to the total variance in the data. Based on this analysis, a subset of components that explains the highest percentage of variance is typically selected. However, multiple valid combinations may exist, and the final choice is often made manually by the researcher. This study introduces a novel yet straightforward algorithm for the automatic selection of the number of principal components. By integrating ANOVA and bootstrapping with principal component analysis (PCA), the proposed method enables automatic component selection in classification tasks. The algorithm is evaluated using three publicly available datasets and applied with both decision tree and support vector machine (SVM) classifiers. Results indicate that this automated procedure not only eliminates researcher bias in selecting components but also improves classification accuracy. Unlike traditional methods, it selects a single optimal combination of principal components without manual intervention, offering a new and efficient approach to PCAbased model development.

Volume: 14

Issue: 3

Page: 1136-1145

Publish at: 2025-12-01

Application of artificial intelligence in emission prediction for hybrid electric vehicles: integrating ANN and GPR

10.12928/telkomnika.v23i6.27128

In recent years, hybrid electric vehicles (HEVs) have emerged as a promising solution to mitigate vehicular emissions and improve fuel efficiency. This study focuses on the Toyota Prius HEV, employing advanced artificial neural networks (ANN) and Gaussian process regression (GPR) to develop a predictive model for vehicle emissions. The model considers multiple pollutants, including carbon monoxide (CO), carbon dioxide (CO₂), hydrocarbons (HC), and nitrogen oxides (NOx), measured under diverse driving conditions. The ANN model predicts emission trends, while GPR estimates prediction uncertainty, enhancing the model’s robustness. The GPR models achieved uncertainty levels of ±0.829 ppm for CO, ±9.978 ppm for HC, ±0.144 ppm for NOx, and ±411.256 ppm for CO₂, respectively, underscoring the robustness of the integrated approach for emission prediction. This research aims to support the development of more sustainable vehicle technologies and inform policy making for environmental sustainability (e.g., Euro 6/Euro 7 standards). Overall, the study addresses how artificial intelligence (AI) can be utilized to achieve accurate multi-pollutant emission predictions in HEVs. The findings reveal that an integrated ANN-GPR approach yields superior predictive performance (R² values approaching 1.0) with quantifiable uncertainty, outperforming a stand-alone ANN model and providing a robust solution to the emission prediction challenge.

Volume: 23

Issue: 6

Page: 1543-1554

Publish at: 2025-12-01

A hybrid ARIMA and DNN approach with residual learning for electric vehicle charging demand forecasting

10.12928/telkomnika.v23i6.27219

Wahyu; National Research and Innovation Agency (BRIN) Cesar , Dwidharma; National Research and Innovation Agency (BRIN) Priyasta , Prasetyo; National Research and Innovation Agency (BRIN) Aji , Melyana; National Research and Innovation Agency (BRIN) Melyana , Agus; National Research and Innovation Agency (BRIN) Suprianto , Osen; National Research and Innovation Agency (BRIN) Fili Nami , Riza; National Research and Innovation Agency (BRIN) Riza

The rapid growth of electric vehicle (EV) adoption has created significant challenges for power grid management and charging infrastructure planning. Accurate forecasting of EV charging demand is therefore essential to ensure reliable electricity supply and effective station deployment. This study proposes a novel hybrid forecasting framework that combines autoregressive integrated moving average (ARIMA) with deep neural networks (DNN) through a residual learning strategy. In this approach, ARIMA models the linear temporal patterns, while DNN captures the nonlinear residuals, resulting in improved efficiency and predictive accuracy. The proposed hybrid model is one of the first applications of the residual learning approach for EV demand forecasting in Indonesia. Experimental evaluation using real-world daily consumption data shows that the hybrid method achieved the highest prediction accuracy of 98.22%, consistently outperforming single-model baselines. Beyond technical performance, the model can support stakeholders in planning charging infrastructure and help maintain grid stability in rapidly growing EV ecosystems.

Volume: 23

Issue: 6

Page: 1555-1565

Publish at: 2025-12-01

A design and reconfigurable phase shift inductor inductor capacitor converter for switch failures

10.12928/telkomnika.v23i6.26926

Xu; Universiti Teknologi MARA Lili , Muhamed Nabil; Universiti Teknologi MARA Hidayat , Nik Hakimi; Universiti Teknologi MARA Bin Nik Ali , Muhammad; Universiti Teknologi MARA Umair

The reliability of a converter operation strongly affects overall system performance and is vital for uninterrupted power-electronic operation. Harsh operating conditions and environmental stresses degrade device performance and reduce reliability. In particular, a switching device failure may prevent an inductor inductor capacitor (LLC) resonant converter from operating near its resonant frequency while still maintaining stable output voltage, potentially causing loss of operation as well as significant drops in both efficiency and power delivery. To address this challenge, this paper proposes a fault-tolerant topology and control strategy for the LLC converter under open circuit switch (OCF) faults. The proposed method integrates a bypass arm with a secondary-side series configuration; when a primary-side open-circuit fault occurs, the auxiliary switch is activated to bypass the faulty leg, reconfiguring the secondary side into a voltage doubler rectifier (VDR). This reconfiguration enables continuous operation with an output voltage doubled relative to the normal condition, while minimizing performance degradation. Simulation results confirm that, even under a single-switch OCF, the proposed approach maintains an efficiency of 98% with output voltage fluctuation limited to less than 1%. Compared to conventional methods, the proposed strategy greatly enhances reliability and fault tolerance, making it well-suited for high-efficiency power conversion applications.

Volume: 23

Issue: 6

Page: 1676-1686

Publish at: 2025-12-01

Reconfigurable ultra-wideband hexagonal antenna with two notched-band features for wireless applications

10.12928/telkomnika.v23i6.27047

Khaled; Azzaytuna University B. Suleiman , Akrem; College of Computer Technology Zawiya Asmeida , Shipun; UTHM University Anuar Hamzah , Mohd Shamian; UTHM University bin Zainal

Owing to the demand for frequency agility, a switchable ultra-wideband (UWB) hexagonal antenna was developed in this study. The proposed antenna features two notch filters introduced by two U-shaped slots on the patch to reduce interference from other wireless networks by rejecting the unique frequency bands. In addition, the proposed antenna comprises a hexagonal radiator attached to a feeding 50 Ω standard microstrip line. To fabricate the antenna prototype, a substrate (Rogers RT/Duroid 5880) with loss tangent and relative permittivity values of 0.0009, and 2.2, respectively, was used. Frequency and pattern reconfigurability were achieved by changing the electrical equivalent circuit of two positive-intrinsic-negative (PIN) diodes sandwiched within two U-shaped slots. The evaluation confirmed that the antenna operated within the D1&D2-ON configuration across the entire UWB range while, effectively filtering the wireless body area network (WBAN) (6.10–6.56 GHz) and radar application (9.16–10.79 GHz) bands when both diodes were OFF. The radiation efficiency and gain reached values of 92.9 % and 7.5 dB, respectively. The proposed design offers a robust performance with enhanced interference rejection. This makes it suitable for modern cognitive radio systems.

Volume: 23

Issue: 6

Page: 1439-1448

Publish at: 2025-12-01

Watermarking on spread multi-frame data video using discrete wavelet transform hybrid and frame ratio message variance

10.12928/telkomnika.v23i6.27211

Ilham; Institut Teknologi Sumatera Firman Ashari , Sarwono; Institut Teknologi Sumatera Sutikno

The exponential growth of video sharing demands secure and imperceptible watermarking methods. This study presents a video watermarking framework using discrete wavelet transform (DWT) with hybrid sub-band embedding and multi-frame allocation to balance imperceptibility, capacity, and robustness. Watermark bits are adaptively distributed across low-low (LL), low-high (LH), and high-low (HL) sub-bands of selected frames, with uncompressed audio video interleave (AVI) ensuring coefficient integrity. Experiments on 640×360 videos show LL-only embedding achieves high imperceptibility (peak signal-to-noise ratio (PSNR) > 38.6 dB, structural similarity index measure (SSIM) ≥ 0.9945, and bit error rate (BER) = 0), while LL-dominant hybrids increase capacity with slight robustness trade-offs. Embedding in LH and high-high (HH) sub-bands raises distortion vulnerability. Under cropping, BER rises from 0.005 to 0.205 (0–50%), and normalized correlation (NC) drops from 0.998 to 0.802, remaining acceptable for ≤30% cropping. The scheme resists joint photographic experts’ group (JPEG) compression quality factor 20–80 (Q20–Q80), resizing (≥70%), and mild Gaussian blur (3×3), maintaining efficient decoding under higher payloads. Future work may apply error-correction coding and redundancy-aware embedding for improved resilience. Overall, the proposed method offers a secure, adaptive, and efficient solution for video authentication and covert communication.

Volume: 23

Issue: 6

Page: 1476-1494

Publish at: 2025-12-01

Dynamic service-aware network selection framework for multi objective optimization in 5G-advanced heterogeneous wireless networks

10.11591/ijai.v14.i6.pp4993-5007

Bhavana Srinivas , Nadig Vijayendra Uma Reddy

The increasing complexity of heterogeneous wireless networks (HWNs) and the diverse requirements of mobility patterns and service classes necessitate advanced solutions for network selection and resource optimization. Existing models often fall short in addressing dynamic mobility scenarios and service differentiation, leading to inefficiencies in resource allocation, suboptimal throughput, and increased latency. To overcome these limitations, this study proposes a dynamic service-aware network selector (DSANS) framework for 5G-advanced environments. The framework integrates an adaptive deep decision network (ADDN) for multi-objective optimization, addressing critical quality of service (QoS) metrics such as throughput, delay, and energy efficiency while enhancing quality of experience (QoE) for applications like enhanced mobile broadband (eMBB), ultra-reliable low latency communication (URLLC), and internet of things (IoT). The DSANS framework dynamically adapts to mobility patterns and varying network conditions, ensuring efficient resource estimation and optimal network selection. Simulation results highlight its superiority, achieving up to 25% improvement in throughput and a 15% reduction in latency compared to state-of-the-art algorithms. These findings validate DSANS as a robust solution for mitigating the limitations of existing models, optimizing network performance, and meeting the stringent demands of next-generation HWNs.

Volume: 14

Issue: 6

Page: 4993-5007

Publish at: 2025-12-01

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

Filters article

Years

Source Title

Authors

31,042 Article Results

An insight on using deep learning algorithm in diagnosing gastritis

Optimizing vehicle inspection efficiency and integrity in Tanzania through blockchain technology

Enhancing handover management in 5G networks with encoder-decoder LSTM for multistep forecasting

Modeling chemical kinetics of geopolymers using physics informed neural network

Particle swarm optimization-optimized integrator backstepping for the control of electric wheelchairs velocity

A blended ensemble approach for accurate human activity recognition

Retrieval-augmented generation for Arabic legal information: the family code case study

Advanced signal transformation techniques to improve spectral efficiency in visible light communication systems

The bootstrap procedure for selecting the number of principal components in PCA

Application of artificial intelligence in emission prediction for hybrid electric vehicles: integrating ANN and GPR

A hybrid ARIMA and DNN approach with residual learning for electric vehicle charging demand forecasting

A design and reconfigurable phase shift inductor inductor capacitor converter for switch failures

Reconfigurable ultra-wideband hexagonal antenna with two notched-band features for wireless applications

Watermarking on spread multi-frame data video using discrete wavelet transform hybrid and frame ratio message variance

Dynamic service-aware network selection framework for multi objective optimization in 5G-advanced heterogeneous wireless networks

Discover Our Library