International Journal on Robotics, Automation and Sciences

Detection of Malicious URLs: A Deep Learning and Machine Learning Perspective

Shougfta Mushtaq — 2026-03-31

Nowadays, as the cyber world is expanding rapidly, issues related to cybersecurity are also increasing. The criminal mind set try to breach the security of individual or organization by firstly, win confidence and secondly attack them for this purpose URL phishing is a most common way where phisher attach a link and share with victim. The proposed paper examines the various machine learning and deep learning approaches on state-of-the-art data set Crawling2024 by classifying the phishing and legitimate URLs. The study involves different machine learning algorithms like Random Forest, LR (Logistic Regression), XGBoost, MLP (Multilayer Perceptron) and Gated Recurrent Units (GRU) and deep learning algorithms like CNN (Convolutional Neural Network), MLP, etc. and analyze performance metrics accuracy, precision, F1 score, Recall, False Positive and False Negative. The RF (Random Forest) classifier achieved the highest precision (98.63%) and accuracy (96.24%), while Logistic Regression and GRU also achieved well. In addition to that LTRCN (Long-Term Recurrent Convolutional Network) achieved good precision but poor accuracy 48.23%. The experimental work shows that conventional algorithms such as Random Forest and advanced algorithms like GRU are efficient in detecting URL phishing, it also emphasizes that there is still need of some advanced approaches like CNN and LTRCN.

Manuscript received: 15 May 2025 | Revised: 26 Jun 2025 | Accepted: 28 Oct 2025 | Published: 31 Mar 2026

An Integrated Controller System For Unmanned Surface Vehicles (USV)

Nazreen Rusli — 2026-03-31

Unmanned Surface Vehicles (USVs) are extensive used in several industries, such as environmental monitoring, offshore resource exploration, and maritime security. The benefits of USV for risk minimization and prolonged operational endurance cause an increase in demand. USVs are essential for administering marine security laws since they can remotely monitor traffic. Their ability to navigate well and avoid collisions improves the efficiency and safety of marine traffic. The incorporation of cutting-edge sensors and battery-powered vehicles enhances the dependability and operating capacities while minimizing environmental impact. Unlike traditional fuel-powered vessels, battery-operated USVs produce no direct water pollutants, contributing to cleaner oceans and more sustainable maritime operations. In response to these technological advancements and the unique maritime needs of Malaysia and neighboring oceanic nations, Centre for Unmanned Technologies (CUTe) at the International Islamic University Malaysia (IIUM) collaborated with Prostrain Technologies to develop a robust controller for USVs called "CxSense". The controller board of CxSense has been designed to meet the stringent compliance requirements of IPx8, ESD test, and vibrations test, demonstrating their robustness, reliability, stability, and adherence to industry standards for high-level protection. This invention has the potential to significantly improve the efficiency and safety of maritime operations in Malaysia and the surrounding oceanic nations.

Manuscript received: 18 Nov 2024 | Revised: 30 Aug 2025 | Accepted: 2 Mar 2026 | Published: 31 Mar 2026

Exploring AI’s Role in Educational Data Mining: A Bibliometric Review of Applications in Learning Analytics

Valentine Kirimi Muriira — 2026-03-31

The paper explores how Artificial Intelligence (AI) is influencing educational data mining (EDM) and learning analytics (LA), tracing the trend of publications from the year 2005 to 2025. The study identifies 1,006 academic articles through co-citation and co-word methods to map the intellectual space of the field using VOSviewer and Scopus. The rise of machine learning, predictive modelling, and personalized learning as a research emphasis has been noted by a significant rise in AI-EDM publications since 2020 that point to the digital learning shift during the pandemic. Analysis of co-citation shows that the EDM work of Romero and Ventura intersects the machine learning work of Breiman, and in this context, the evolution of explainable AI, predictive analytics, and early intervention techniques. Co-word analysis reveals three main themes, such as students, data mining, and machine learning, which characterize AI-based educational practices. Thematic clusters refer to the increased attention to adaptive systems, student performance prediction, and AI ethics. The study highlights how a combined strategy of real-time education and ethical teaching methods can assist AI in helping customize learning and inclusivity among students. Nevertheless, the Scopus database and publications in the English language are the limitations of the study, which can omit the global innovations. Future studies are supposed to promote multidisciplinary teams and create emotional AI algorithms with high ethical codes to overcome privacy and equality issues. This paper offers a detailed evaluation of AI as an agent of change in education and will be useful to decision-makers interested in building equitable and data-driven educational systems.

Manuscript received: 24 Apr 2025 | Revised: 26 Feb 2026 | Accepted: 3 Mar 2026 | Published: 31 Mar 2026

The Impact of AI on Educational Content Creation: A Bibliometric Review of Trends, Influential Research, and Emerging Themes in Enhancing Access and Quality in Education

Valentine Kirimi Muriira — 2026-03-31

The research conducts a detailed examination of Artificial Intelligence usage in educational content development to identify its effect on teaching methods while making connections to SDG 4 about quality education. The bibliometric dataset comprised of 494 peer-reviewed publications from Scopus database (2005–2025) was used to conduct co-citation, co-word and trend analysis through VOSviewer for identifying intellectual structures together with thematic clusters as well as emerging research trajectories. The analysis shows a major growth in educational AI research following 2020 because institutions adopt adaptive learning systems and generate content automatically while using intelligent teaching methods. Self-regulated learning along with predictive analytics data mining and immersive learning environments and self-regulated learning formed the four strongest co-citation groups. A co-word analysis confirmed that the core words “e-learning,” “learning systems” and “machine learning” function as fundamental components within this academic domain. The study outlines the main ethical and practical consequences which include algorithm-based bias together with data administration problems and unequal access to materials that that emphasize the value of developing AI systems consider diverse contexts. The paper presents practical guidelines for leaders in research and education institutions alongside government agencies which direct AI utilization toward productive results while supporting equality and academic integrity across all programs. The analysis presents a strategic plan that guides how to use AI innovation to develop inclusive educational systems which produce meaningful learning outcomes while sustaining their operations.

Manuscript received: 25 Apr 2025 | Revised: 2 Mar 2026 | Accepted: 4 Mar 2026 |Published: 31 Mar 2026

Integrating Real-Time Pose Estimation in Block-Based Programming Environments Through Novel Architectural Patterns

Kai Liang Lew — 2026-03-31

This technical demonstration study develops an automated motion analysis system via MitApp Inventor. It is a high-level block-based visual programming language. The system also utilises a pose estimation library for computer vision tasks, which is implemented within the application. The system addresses the growing need for accessible motion tracking by eliminating dependency on additional hardware and providing real-time movement classification capabilities. The user interface, as well as the block diagram of the application, are designed and developed using MIT App Inventor. The basic working principle of how the application operates is that users can perform movements that are automatically tracked and classified. MitApp Inventor allows users to design and develop via a computer or a laptop. Once created, the application can be viewed in an Android / iOS emulator as well as on the user's device. In terms of motion tracking performance, Posenet has been chosen as the only library that the MitApp Inventor supports. The Posenet model is suitable for detecting and tracking key body points of a human body in real-time. The system features four different arm exercises, including left-arm bicep curls, right-arm bicep curls, lateral raises, and military presses. These exercises are designed to detect the angles of the body's joints when a user performs them. Testing with 10 participants, who performed 25 repetitions of each exercise, totalling 1,000 pose classifications, demonstrated the system's effectiveness. The Posenet achieved high accuracy in movement recognition, with precision and recall values of 0.94 and 0.94 for left arm curls, 0.932 and 0.932 for right arm curls, and 0.96 and 0.96 for both lateral raises and military push exercises, demonstrating its effectiveness in precise motion classification. The system achieved an overall accuracy of 94.8% while providing immediate feedback for movement form correction, offering a viable approach for automated motion analysis applicable to human-robot interaction, motion capture systems and industrial safety monitoring.

Manuscript received: 9 Jun 2025 | Revised: 17 Jul 2025 | Accepted: 24 Sep 2025 | Published: 31 Mar 2026

Big Data Analytics in Digital Banking Fraud Detection Technologies and Methods

Kai Liang Lew — 2026-03-31

Digital banking fraud has escalated dramatically with the proliferation of online financial services, causing billions in annual losses and threatening the stability of global economic systems. This paper examines the Big Data Analytics (BDA) technologies and methods for real-time fraud detection in digital banking environments. The paper discusses the evolution from traditional rule-based systems to modern distributed computing frameworks, analysing how Apache Hadoop and Spark enable the processing of massive transaction volumes with varying trade-offs between latency and accuracy. Key machine learning approaches are covered, including supervised methods, unsupervised methods, and hybrid architectures that combine both paradigms. The paper identifies critical implementation challenges across technical dimensions, operational aspects, and regulatory requirements. Emerging trends explored include federated learning for privacy-preserving model training, blockchain integration for cross-institutional fraud detection, and edge computing for ultra-low latency inference. The analysis shows that while individual studies report improvements in detection, challenges remain in real-world validation, model interpretability, and cross-institutional generalizability. The paper concludes with practical recommendations for implementing hybrid streaming-batch architectures, embedding explainability mechanisms, and adopting privacy-preserving techniques. This paper provides insight for researchers and practitioners to understand the current capabilities, limitations, and future trends of BDA in enhancing fraud detection in increasingly complex digital banking ecosystems.

Manuscript received: 9 Jun 2025 | Revised: 6 Aug 2025 | Accepted: 24 Sep 2025 | Published: 31 Mar 2026

Big Data Analytics Technologies and Implementation Challenges for Smart Infrastructure Development in Malaysia: A Narrative Review

Chean Khim Toa — 2026-03-31

This paper reviews Big Data Analytics (BDA) applications for traffic management and security monitoring in Malaysian urban infrastructure, examining implementation barriers and scaling requirements for effective adoption. Through an analysis of international case studies from Singapore, São Paulo, and comparable cities, the review identifies four primary barriers preventing the scaling of integrated BDA in Malaysian cities: technological infrastructure gaps, governance fragmentation, regulatory inadequacies, and implementation capacity constraints. Despite recent legislative developments, including the Data Sharing Act 2025 and existing BDA implementations such as Malaysia City Brain, implementation barriers persist due to Malaysia's federal governance structure and infrastructure limitations, which differ significantly from those of successful city-state implementations. This study proposes a contextualised five-priority implementation approach that emphasises expanding existing corridor-based pilots, leveraging federal coordination mechanisms through the Data Sharing Act 2025, regulatory development, strategic partnerships, and performance measurement systems. The analysis reveals that governance coordination and the development of a regulatory framework should guide the scaling of infrastructure investments for effective BDA adoption. These findings provide Malaysian policymakers with a realistic scaling pathway that builds upon existing implementations, working within institutional and resource constraints to develop integrated urban analytics systems for traffic optimisation and security monitoring.

Manuscript received: 9 Jun 2025 | Revised: 17 July 2025 | Accepted: 24 Sep 2025 | Published: 31 Mar 2026

A Narrative Review on Big Data and Social Media Behaviour Analysis for Crisis Response in Thailand During COVID-19 and Flooding Events

Chean Khim Toa — 2026-03-31

Social media platforms have evolved into critical real-time information hubs during crisis events, generating massive user-generated content that presents both opportunities and challenges for emergency management. Thailand's experience with the COVID-19 pandemic and recurring flood disasters provides valuable insights into leveraging social media big data for crisis response in developing countries. This narrative review synthesizes existing literature on technological frameworks, analytical methods, and practical implementations through comprehensive analysis of published studies and documented case studies. By examining distributed computing platforms, natural language processing, sentiment analysis, and geospatial mapping, this review assesses how Thailand has utilised user-generated content for emergency management. The findings reveal both technological progress and persistent systemic constraints. While initiatives such as the Anti-Fake News Centre demonstrate effective misinformation detection within two hours, significant gaps remain in five key areas, including technological infrastructure fragmentation among 48 disaster management agencies, analytical limitations in Thai-language processing, governance framework deficiencies, stakeholder coordination constraints, and digital inclusivity challenges that exclude vulnerable populations. Despite technological implementations, critical barriers include 96% failure rates in monitoring equipment and limited real-time data integration. The analysis provides a systematic examination of implementation gaps spanning technological, analytical, governance, stakeholder coordination, and inclusivity dimensions while identifying strategic opportunities, including enhanced data quality frameworks, cloud-based scalability solutions, and explainable AI integration, to strengthen Thailand's digital crisis management capabilities.

Manuscript received: 10 Jun 2025 | Revised: 7 Aug 2025 | Accepted: 24 Sep 2025 | Published: 31 Mar 2026

Scratch Train for Lightweight Models for Face Mask Detection

Kai Liang Lew — 2026-03-31

Automated systems for detecting face mask use in public became urgent during the COVID-19 pandemic. Most existing mask detection research fine-tunes ImageNet pre-trained backbones on relatively small datasets of masks. This approach raises concerns about model performance in situations with limited computational resources or when external pre-trained weights are not accessible. Additionally, there is a limited comparative analysis of recent lightweight architectures under consistent training conditions for mask detection tasks. This paper evaluates four state-of-the-art lightweight architectures for binary mask detection, including RepViT, ShuffleNetV2, EdgeNeXt-Small, and EfficientFormer. These models are trained from scratch using identical training protocols on two datasets containing 7,553 and 12,000 RGB images, respectively. The performances are then assessed using standardised metrics, including accuracy, precision, recall, and F1-score. Results demonstrate that ShuffleNet V2 achieves an optimal balance between classification accuracy and computational efficiency, delivering 0.987 accuracy on Dataset 1 and 0.998 accuracy on Dataset 2 while maintaining the fastest inference time of 0.464-0.667 milliseconds and the smallest model size of 1.26 million parameters. RepViT and EdgeNeXt-Small achieve slightly higher accuracy but require significant computational resources. EfficientFormer consistently underperforms across all evaluation metrics. These findings indicate that extremely lightweight CNNs can excel at mask detection when trained from scratch, making ShuffleNet V2 the ideal choice for resource-constrained deployment scenarios.

Manuscript received: 12 Jun 2025 | Revised: 30 Jul 2025 | Accepted: 24 Sep 2025 | Published: 31 Mar 2026

Bridging the Communication Gap: A Flexible Sensor-Based Smart Glove for the Speech-Impaired with Custom Phrase Functionality

Fitra Nurakbar — 2026-03-31

In daily life, we frequently encounter individuals with restricted verbal communication abilities, commonly referred to as having a speech impairment. They utilize hand and finger gestures known as ASL (American Sign Language) to communicate with others, including those with speech difficulties and individuals without such challenges. Nevertheless, numerous ordinary individuals lack comprehension of this mode of communication. In this technological era, the author has developed a smart hand glove translator capable of converting speech-impaired hand motions into comprehensible sounds. The primary elements of this design are the Arduino Uno as the central data processing unit, complemented by a flex sensor and a gyroscope sensor to capture hand movements. Subsequently, employing an LCD and speaker as output devices, along with a micro SD reader to access a pre-existing database, facilitates user interaction in daily activities. The ultimate outcome of this design is an intelligent glove equipped with 50 databases, accessible to users through the selection of required phrases.

Manuscript received: 31 Aug 2025 | Revised: 16 Sep 2025 | Accepted: 2 Mar 2026 | Published: 31 Mar 2026

Impact of Image Pre-Processing and Optimizers on U-Net Segmentation of Prostate Cancer in MRI

Anis Amira Abdul Aziz — 2026-03-31

Prostate cancer remains among the most frequently diagnosed cancers in men and is a major issue in achieving timely and accurate diagnosis. Magnetic Resonance Imaging (MRI) is widely applied because it is capable of picking up high-resolution anatomical information, but cancer regions must be manually segmented, which is time-consuming and prone to the variability of experts. This study proposes an automated segmentation algorithm using the U-Net deep learning model with image pre-processing techniques to address these deficiencies. Median filtering was done to remove salt-and-pepper noise, followed by brightness enhancement to improve the intensity contrast of images. The pre-processed images were used to train a U-Net model for prostate cancer segmentation. The Dice Similarity Coefficient (DSC) metric was used to evaluate the segmentation accuracy. Three optimizers, Adam, RMSprop, and Adagrad, were tested. All of them were trained between 10 and 100 epochs. Adam optimizer obtained the highest segmentation performance at epoch 90 with the highest DSC value of 0.9907, while RMSprop and Adagrad produced 0.9888 and 0.9655, respectively. Pre-processing raised the mean DSC from 0.8206 to 0.8733, confirming its impact on image quality enhancement. Overall, the proposed method demonstrates high accuracy and reliability, offering a practical solution to support radiologists in prostate cancer diagnosis and treatment planning.

Manuscript received: 9 Oct 2025 | Revised: 5 Dec 2025 | Accepted: 26 Feb 2026 | Published: 31 Mar 2026

Toward a Smarter EV Infrastructure: Patterns, Gaps, and Opportunities in Public Charging Station Networks

Fikri Mahendra — 2026-03-31

The rapid adaptation of electric vehicles (EV) requires a reliable and easy -to -access public load infrastructure. In order to find the smarter trends, gaps and opportunities for infrastructure development, this study considered the public charges of electric vehicles around the world. We study chargers, costs, space access and distribution measures using a data set of 5,000 charging stations. To optimize the position of the charger and attract attention to the difference in access between urban and rural areas, mixed programming and statistical analysis are used. The results show that urban areas have a higher load but are blocked, while rural areas do not have enough coverage. Improve cybersecurity procedures and integrate renewable energy sources are two areas that can be used improved. The results are supported by recent documents, emphasizing the importance of the fair and effective EV infrastructure plan.

Manuscript received: 16 Jun 2025 | Revised: 5 Aug 2025 | Accepted: 13 Nov 2025 | Published: 31 Mar 2026

Techno-Economic Analysis of Renewable Energy Trends Based on Historical Data

Ahmad Suharjo Marinda — 2026-03-31

The desire to lessen reliance on fossil fuels and combat climate change has been cited as the reason for the recent global increase in the usage of renewable technologies. This research aims to give a comprehensive techno-economic evaluation of solar and wind energy for the years 2009–2021, using four key datasets: average prices, installed capacity, and generating numbers. Python was used for data processing and visualization, and trends in energy output, capacity expansion by source, and expenses over time were analyzed. Solar and wind capacity increased dramatically between 2014 and 2020, rising from 44.56 GW to 462.44 GW and from 95.88 GW to 209.14 GW, respectively. The cost of generating has also dropped significantly: wind power now costs $29.28/MWh instead of $74.33/MWh, while solar energy now costs $23.13/MWh instead of $167.67/MWh. Also noted were seasonal changes in energy output and an uptick in demand for storage. The results strongly support the central position of solar and wind resources for a sustainable energy system and refine policy and investment guidance rationale. Addressing the remaining data gaps and the particular difficulties in growing these energy sources should be the top priorities of future initiatives.

Manuscript received: 16 Jun 2025 | Revised: 22 Aug 2025 | Accepted: 2 Mar 2026 | Published: 31 Mar 2026