International Journal of Internet Manufacturing and Services (32 papers in press)

Regular Issues

• Open Access

  An Adaptive Variable Frequency Control Method for Motor Speed Based on PLC ( Free Full-text Access )

  by Bingyu Wang 
  Abstract: The traditional fixed frequency motor drive method is no longer able to meet the response requirements of dynamic loads under complex working conditions, resulting in high steady-state errors and overshoot in motor speed variable frequency control. Therefore, a PLC based motor speed adaptive variable frequency control method is proposed. Firstly, an efficient and stable motor speed control architecture was constructed using the USS protocol and RS485 interface. Secondly, the timing interrupt mechanism and high-speed counter are used to collect the pulse signal of the motor encoder in real time and calculate the speed. Finally, based on the fuzzy PID control strategy, adaptive variable frequency control of motor speed is achieved. The experimental results show that the method proposed in this paper has almost zero error in the steady-state error test of motor speed, and the highest overshoot suppression rate reaches 97.34%.
  Keywords: PLC technology; Motor speed; Time base interruption; Fuzzy PID; Frequency conversion control.
  DOI: 10.1504/IJIMS.2026.10079581
   
  
• How Algorithmic Leadership Profiles Influence Employee Task Performance and Workplace Well-being: a Typological Approach  
  by Qiao Chang, Zhang Jian, Xiao Xiao, Shahan Bin Tariq 
  Abstract: Developing knowledge about algorithmic leadership effectiveness requires clarifying its behavioural manifestations. However, current research overly emphasises a single form algorithmic control while neglecting other manifestations. Drawing on leadership behaviour theory and research on algorithms impacts on employee task performance and workplace well-being, this study proposes two fundamental dimensions: algorithmic control and algorithmic norm. Using a typological approach, we construct a four-type model (high control-high norm, high control-low norm, low control-high norm, low control-low norm). A pilot study (N = 779) validates these dimensions, and the main study (N = 137) reveals that high-control, high-norm algorithmic leadership significantly enhances employees task performance and workplace well-being. Theoretically, this typology expands understanding beyond a human-oriented perspective; practically, it guides organisational implementation of algorithmic leadership.
  Keywords: Algorithmic Leadership; Typology; Task Performance; Workplace Well-being.
  DOI: 10.1504/IJIMS.2026.10070367
   
  
• Eco-Labels and E-Commerce: exploring Consumer Purchasing Behaviour Through Digital Literacy, Credibility, and Emotions  
  by Shaema Ali Abdullah Al-Olfi, Yazeed Al-Hajj, Yanan Song 
  Abstract: This research investigates how eco-labels influence e-commerce consumer purchasing behavior, focusing on the mediating roles of digital literacy and information credibility, as well as the moderating effects of consumer guilt and self-efficacy. A total of 417 valid responses were collected from 500 distributed questionnaires through convenience sampling on Google Forms and Wenjuanxing. The survey was offered in Chinese, English, and Arabic to ensure inclusivity. Data analysis utilized Smart PLS 4.1.0 software for PLS-SEM. Findings reveal that eco-labels significantly affect purchasing behavior by enhancing digital literacy and information credibility, both acting as partial mediators. Consumer guilt emerged as a key moderator, while self-efficacy showed no significant moderating effect. This study fills a gap in the literature by examining the indirect effects of eco-labels through digital literacy and information credibility. It also emphasizes the importance of emotional factors, particularly consumer guilt, in developing effective eco-labeling strategies for sustainable purchasing.
  Keywords: Eco-labels; E-commerce ; Consumer purchasing behavior; Digital literacy; Information credibility; Consumer guilt; Self-efficacy.
  DOI: 10.1504/IJIMS.2026.10070741
   
  
• Enhancing Security Mechanisms in Cloud Computing Optimisation for Modern Data Protection Strategies  
  by K. Hazeena, C.Sathish Kumar, G. Prathap, S.Silvia Priscila, Edwin Shalom Soji, S. Suman Rajest 
  Abstract: This study explores the latest advancements in security mechanisms designed for cloud computing environments, addressing the urgent need for robust protection in an evolving technological landscape. By leveraging advanced encryption techniques, blockchain technology, and AI-driven anomaly detection, the study assesses the effectiveness of these solutions in protecting sensitive data from modern cyber threats. Data were collected from various cloud service providers, including public, private, and hybrid cloud models. The research utilized tools such as AWS CloudTrail for monitoring, IBM's QRadar for threat intelligence, and TensorFlow for developing AI algorithms to perform comprehensive security assessments. The findings reveal significant improvements in threat detection accuracy, faster response times, and enhanced data integrity. The paper concludes with recommendations for integrating these advanced security measures to strengthen cloud infrastructure against potential breaches. These insights aim to assist cloud service providers and users in adopting effective strategies to enhance their security posture in the face of evolving threats. By adopting these cutting-edge solutions, organisations can better safeguard their data, ensuring greater resilience and trust in their cloud computing environments.
  Keywords: Cloud Security; Encryption Techniques; Blockchain Technology; AI-driven Anomaly Detection; Cyber Threats; Cloud Service Providers; Threat Intelligence; Data Integrity.
  DOI: 10.1504/IJIMS.2027.10071761
   
  
• Investigation of Microstructure to Research the Metallurgical Characteristics of Recently Manufactured Mg-HAP Biodegradable MMCs  
  by Neeraj Kumar, R.A.J. Kumar Duhan 
  Abstract: Due to their biocompatibility and resemblance to actual bone, magnesium alloys are a potential material for orthopaedic implants. Matrix composites offer a means to improve the material's metallurgical properties since magnesium corrodes too quickly in the body's environment. Within this current investigation, stir casting utilised to generate Mg-HAP MMCs. One type of reinforcing material that has been employed is HAP powder. Following casting, composites were formed into the necessary form samples for testing by metallurgists. Utilising energy dispersive X-ray spectroscopy (EDXS), optical microscopy (OM), elemental composition evaluation and scanning electron microscopy (SEM), the microstructure of the magnesium-HAP composite was investigated. Grains in Mg-HAP composites get smaller when HAP is included in the magnesium matrix. Comparisons are made between the composites ultimate tensile and compressive strengths when HAP is added. It has been discovered that the inclusion of HAP reduces the ultimate tensile and compressive strengths of composites.
  Keywords: Biodegradable; Magnesium Hydroxyapatite; Metal Matrix Composites; Stir Casting; Ultimate Tensile Strength; Compressive Strength.
  DOI: 10.1504/IJIMS.2027.10072560
   
  
• The Mediating Mechanism of Bank Efficiency in the Nexus between Liquidity, Credit Risks, and Bank Performance: a Comparative Study  
  by Alaa Omran, Xiao Ming, Ibrahim M. Menshawy, Ahmed Hamdy 
  Abstract: This study investigates the influence of credit and liquidity risks on the performance of Islamic and conventional banks in the Middle East, with particular emphasis on the mediating role of bank efficiency in this relationship over the period from 2008 to 2023. Based on a sample comprising 70 conventional banks and 12 Islamic banks in the Middle East -representing 1,120 and 192 bank-year observations respectively- the outcomes show that Islamic banks have a higher average efficiency than conventional banks, despite having a similar return on assets. Furthermore, liquidity risk indirectly enhances bank performance through bank efficiency. Compared to conventional banks, Islamic banks exhibit a significant positive association between credit and liquidity risks and bank efficiency. Moreover, efficiency plays a key role in enhancing performance in Islamic banks. In contrast, in conventional banks, credit and liquidity risks tend to reduce efficiency, although efficiency still has a positive influence on performance. However, in both Islamic and conventional banks, credit and liquidity risks negatively affect overall bank performance. Finally, this investigation provides significant theoretical and practical insights into banking performance and risk management.
  Keywords: Bank Performance; Credit Risk; Bank Efficiency; Liquidity Risk; conventional banks; Middle East banks; Islamic Banks.
  DOI: 10.1504/IJIMS.2027.10072744
   
  
• A Hybrid Deep Learning Approach with Interpretability on BiLSTM-Dense SHAP Enhanced Model (BDSEM) for Depression Prediction  
  by S.Nalini Poornima, S. Geetha 
  Abstract: An increasing number of individuals have experienced anxiety and despair in recent years. These symptoms may be quite harmful and are difficult to identify. The most severe forms of depression, a worldwide mental health issue, may result in suicide or self-harm. A person's everyday life is negatively impacted by depression and other associated psychiatric diseases, which are becoming more prevalent. People's behaviour and mental health need to be monitored immediately, and automated software that runs in the background can assist in tracking this status with high accuracy and tolerance. The ability to evaluate the components influencing the incidence and clinical presentation of depression has led to the advancement of machine learning (ML) frameworks for automated diagnosis of depression in recent years, reaching the next level of deep learning frameworks. An automated depression detection system greatly aids clinical diagnosis and early depression treatments. To forecast depression, we propose a BiLSTM-Dense SHAP enhanced model (BDSEM) system in this study, which combines sophisticated deep learning methods with the interpretability of the model. The bidirectional long short-term memory (BiLSTM) network is the primary component of the BDSEM system. Our approach significantly enhances both efficiency and detection accuracy.
  Keywords: Depression prediction; CNN and RF; LSTM and SHAP; Bidirectional Long Short-Term Memory (BiLSTM); Psychological Disorders; Mental Health; Anxiety and Despair.
  DOI: 10.1504/IJIMS.2027.10072773
   
  
• Harnessing IoT Solutions for Enhanced Electric Power Management  
  by Swaminathan Ramamurthy, S. Saranya, Vaidianathan Balasubramanian, Thejo Lakshmi Gudipalli, K. Arun, J. Jayaprakash, M. Mohamed Sameer Ali 
  Abstract: The purpose of this paper is to research the integration of IoT solutions into the electric power sector with a focus on their transformational potential in optimising the power distribution and consumption. By using technologies such as smart metres, sensors, and automatic control systems, power networks become efficient, reliable, and sustainable. This system enables real-time data collection and analysis, allowing insights into the operations of the grid and improving load management. Analytics engines, AI-based decision systems, and cloud data storage provide predictive maintenance, demand response, and distributed resource energy management, reducing energy loss and operational costs. Other subtopics include IoT contributions to smart grids and renewable energy integration. The Automated Demand Response (ADR) project by Tata Power Delhi Distribution Limited proves that IoT may improve grid efficiency. However, data security concerns, high implementation costs, and high-bandwidth infrastructure must be overcome to maximise its potential. Case studies and empirical evidence suggest that IoT-based solutions can help achieve global energy efficiency targets and ensure responsive and sustainable power systems. IoT is crucial to energy management and a sustainable energy ecology, according to this study.
  Keywords: IoT Solutions; Electric Power Management; Smart Grid Technology; Predictive Maintenance; Energy Efficiency; Renewable Energy Integration; Smart Meters; Data Security; Sustainable Power Networks.
  DOI: 10.1504/IJIMS.2027.10072838
   
  
• An Intelligent Approach with Integrated Feature Selection for Accurate Parkinson's Disease Diagnosis  
  by Pravin Madhukar Dhanrao, B.K. Madhavi, Haard Patel, Debabrata Swain, Sashikala Mishra, Het Prajapati, Shrey Joshi 
  Abstract: Parkinson disease (PD) is a progressive disease which causes slow movement, tremors, imbalance, pain, sensory disturbances, and mental health disorders. There have been methods that were proposed to help in early detection of Parkinson Disease. These methods include analyzing speech data, gait patterns, smell identification and many more. There is no permanent treatment developed for PD. But if the disease can be detected at an early stage, then its further progression can be delayed through proper medication. This paper explores Machine learning techniques like SVM, Adaboost and Voting classifier for detection of PD. Dataset used here contains 22 input features those represent acoustic parameters extracted from speech data. During data preprocessing feature scaling, data balancing operations and feature reduction are performed for enhancing the performance. For performance boosting of the models hyperparameters are tuned using grid searchCV. During model validation it is found that SVM outperforms other classifiers with an accuracy of 98.3%, while Adaboost and Voting classifiers have shown an accuracy of 96.61% and 94.41% respectively.
  Keywords: Parkinson’s Disease; Machine Learning; SVM; Adaboost; Voting Classifier; Normalization; PCA.
  DOI: 10.1504/IJIMS.2027.10072985
   
  
• Nutrient Deficiency Classification using D-ResNet50 for Rice Crop in Indoor Farming Systems  
  by D. Radha, S. Prasanna 
  Abstract: The cultivation of rice is a well-known industry in India. Rice, one of the most important crops and a significant influence on people's health, is cultivated by a large number of people as a means of subsistence. It generates job opportunities, and the production of rice is critical to a large number of small businesses. Nutrients are essential for ensuring the production of the desired crop and maintaining its quality. It is essential to precisely detect nutrient deficiencies in plants to accomplish the goal of providing plants with the appropriate quantities of fertilizer. The physical examination of symptoms and the identification of vitamin deficiencies require a great deal of expertise. Considering that these deficiencies may lead to various health problems, it is essential to identify and address them as soon as possible. We developed a hybrid model for categorising rice plants deficient in potassium (K), phosphorus (P), and nitrogen (N) in this study, based on our findings. The DenseNet121 and ResNet50 networks serve as the foundation for the hybrid model that represents NPK nutritional deficiencies in rice crops. In terms of accuracy and performance, we can conclude that our proposed model outperforms existing models.
  Keywords: Classification Accuracy; Nutrient Deficiencies; DenseNet121 and ResNet50; Indoor Farming Systems; Environmental Conditions; Remote Sensing; Crop Health Monitoring.
  DOI: 10.1504/IJIMS.2027.10073261
   
  
• Optimal Scheduling of Digital Product Innovation: a Case Study of Yonyou  
  by Liyang Wang, Min Zhang, Hongrui Liang, Zi Wang, Botong Liu 
  Abstract: Digital product innovation is crucial for firms' competitive advantages in digitalisation. This process is inherently multi-staged with continuously new module introduction beyond existing scope in response to market demand. While methods for decision-making and planning in digital product innovation are provided, the scheduling of concrete module introduction for effective innovation remains underexplored. We formulated an optimal scheduling model based on a digital product: Yonyou U8 software. The model, with the optimisation goal of minimising innovation cost for a timely market response, aims to output a reasonable module introduction sequence under constraints on capability resources and man-hours. The goal is fully utilisation of capability resources with the least idle man-hour time. Using an applied mixed-integer programming algorithm, it has been proven effective through empirical data. We identify key decision variables and provide methods for planning product innovation actions. It offers practical value for similar firms to to conduct digital product innovation.
  Keywords: Digital product innovation; Optimal scheduling; Case study; Response time.
  DOI: 10.1504/IJIMS.2027.10073903
   
  
• Semi-Supervised Skin Lesion Segmentation Using Structured Prediction-Based Deep Reinforcement Learning  
  by M. Megala, T.R. Nisha Dayana 
  Abstract: Automatic skin lesion segmentation algorithms are vital for reducing dermatologists' workload and forming the basis for early cancer detection. However, the scarcity of large annotated datasets in the medical field limits the performance of deep learning models, often leading to overfitting. The need for vast annotated datasets to train deep models for skin lesion segmentation is impractical in medical scenarios, intensifying overfitting problems. To overcome these challenges, we propose a semi-supervised training method that utilises a Structured Prediction-based Deep Deterministic Policy Gradient, which leverages unlabelled data to enhance neural network robustness. This study presents a novel two-stage training framework for skin lesion segmentation. In the first stage, a supervised approach is used to learn semantic segmentation maps, followed by an unsupervised strategy in the second stage to refine the Structured Prediction-based Deep Deterministic Policy Gradient, thus reducing dependency on large annotated datasets. By incorporating unlabelled samples, we enhance the model's generalisation capabilities. Unlike traditional skin lesion segmentation methods, our approach introduces a surrogate task that leverages convolutional and transformer representations to extract data-driven features from images, thereby minimising the need for extensive annotations. Our method shows promising results across three different skin lesion segmentation datasets: ISIC 2018, ISIC 2017, and PH2.
  Keywords: Convolutional Neural Networks; Data Augmentation Techniques; Medical Image Processing; Multi-Task Learning; Skin Cancer Detection; Supervised Learning; Unsupervised Learning; U-Net Architecture.
  DOI: 10.1504/IJIMS.2027.10073920
   
  
• Feature Selection and Ensemble Classification System for Depression Detection from Social Media Data  
  by S. Saranya, G. Usha 
  Abstract: Without early treatment and depression-detecting applications, hundreds of millions of individuals have mental problems. Depression, one of the main causes of anxiety, bipolar disorder, and sleep disorders, can lead to self-harm and suicide attempts. In healthcare, media, and e-commerce, social media data is increasingly used to detect depression quickly. Twitter, Facebook, and Weibos big and complicated data can reveal peoples emotional and psychological states. However, selecting the relevant features from large and complex datasets is challenging. Depression detection using feature selection approaches and an ensemble DL classifier is computationally feasible in this paper. NLP-based feature extraction, random forest and XGBoost feature selection, and deep neural networks (DNNs) and CNN algorithms are used. The method employed benchmark data from Twitter concerning a dataset of depression-related tweets, including depression detection information. Python, TensorFlow, Keras, and Scikit-learn are used to train and test the suggested technique on the dataset. Experimental results show that the ensemble classifier outperforms machine learning-based emotion identification algorithms in terms of accuracy, precision, recall, and F1-score. Real-time depression identification is possible using the proposed approach.
  Keywords: Social Media Data; Depression detection; System Security; Feature Selection; Deep Learning; Machine Learning; Ensemble classifier.
  DOI: 10.1504/IJIMS.2027.10074365
   
  
• Hybrid AES-Twofish for Data Privacy Supports Knowledge Process Enterprise Application Cloud Migration Optimisation  
  by Karthikeyan Sivanandi 
  Abstract: Cloud migration is essential for organisations seeking to use cloud platforms scalability and flexibility in knowledge management. Optimising this migration process while protecting data is difficult. This article presents an optimised cloud migration architecture utilising a Stackelberg game model and a hybrid encryption strategy that combines AES and Twofish algorithms to enhance data privacy. The cloud service provider sets pricing and quality of service strategies, while the organisation follows and migrates based on these factors in this game-theoretic model. A Stackelberg game equilibrium is reached when the organisation maximises benefits and the cloud provider optimises resource allocation. The hybrid AES-Twofish encryption approach protects important company data during migration by combining AESs efficiency and Twofishs security with low performance impact. The proposed approach offers an optimised, cost-effective cloud migration procedure and handles data security and privacy concerns, making it ideal for organisations with strict data protection standards. Simulation results indicate that the proposed strategy enhances migration efficiency, reduces costs, and improves data security. Combining game theory and cryptography, this research provides a comprehensive solution for organisations seeking to optimise their cloud migration while safeguarding vital data.
  Keywords: Cloud Migration; Knowledge Management Process; Stackelberg Game; Data Privacy; AES-Twofish Encryption; Optimization; Enterprise Application Cloud.
  DOI: 10.1504/IJIMS.2027.10074394
   
  
• Exploring the Association Between Nutrient Intake Level Children Prediction Based on Soft Max Generative Adversarial Networks  
  by P. Jeyanthi, R. Durga 
  Abstract: Deep learning uses neural networks with several layers to look at complicated data, including how children’s health is affected by their eating habits. Food insecurity and insufficient micronutrient consumption can significantly impair the physical, mental, and emotional development of children aged six months to five years. This research examines the capacity of deep learning to evaluate and forecast dietary deficits utilising various datasets. There are five main steps in developing the model: pre-processing with linear min-max normalisation to get rid of missing or irrelevant data; spectral support to check the quality of the data; and feature selection with spectral RFE to get rid of low-weight features. A sigmoid activation function is used to forecast the probability of nutrient intake, and it limits values to [0, 1] for reliable binary classification. Lastly, the improved weighted SmGAN looks at text-based nutritional data to predict food insecurity risks and changes in sentiment. The suggested SmGAN framework is more accurate at making predictions than typical categorisation models. This makes it a strong tool for enhancing the assessment of child nutrition.
  Keywords: deep learning; young children; nutrient intakes; insecurity; recursive feature elimination; normalisation; softmax generative adversarial network; SmGAN; classification and feature evaluation.
  DOI: 10.1504/IJIMS.2027.10074397
   
  
• Advanced K-Nearest Neighbour Models with Meta-Heuristic Algorithms Improve California Bearing Ratio Predictions  
  by Huan Huang, Long Lei, Guoxin Xu, Jiantao Dang, Xiping Ren 
  Abstract: The California bearing ratio (CBR) is a crucial metric in geotechnical engineering that is used to assess the strength of soil. The K-nearest neighbour (KNN) method is a crucial modelling methodology designed to achieve the maximum degree of accuracy in forecasting CBR values. The goal of this work is to develop a prediction model that utilises the KNN approach to estimate CBR while accounting for key input factors, including soil quality, compaction characteristics, and moisture content. For training and testing, this study utilised a dataset of 109 soil samples, enabling a performance comparison with traditional regression-based models developed for the same purpose. The combination of two meta-heuristic techniques, the dandelion optimiser (DO) and the Dingo optimisation algorithm (DOA), to improve the models accuracy, is what distinguishes this work. The studys conclusions identify three distinct models: the KNN, KNDA, and KNDO models. Each of these models provides useful data to ensure precise CBR projections. With an astounding R2 score of 0.987 and an incredibly low root mean square error (RMSE) value of 1.251, the KNDO model is clearly the best performer.
  Keywords: California Bearing Ratio; K-Nearest Neighbors; Dandelion Optimizer; Dingo Optimization Algorithm; Dingo Optimization; Predictive Models; Flexibility and Dispersion.
  DOI: 10.1504/IJIMS.2027.10074592
   
  
• Extraction and Matching of Finger Vein and Iris pattern-based Authentication System using CNN and Deep Learning  
  by K. Appasamy, R.S. Shanmugasundaram 
  Abstract: Biometric authentication systems have recently overcome traditional systems through the application of deep learning to increase soundness and accuracy. This paper proposes an integrated biometric verification system that combines finger vein and iris texture highlight features using convolutional neural networks (CNNs) and deep learning systems. The system consists of finger vein and iris texture unique features that are realistically impossible to reproduce or counterfeit. Feature extraction is achieved through a tailored CNN structure trained on pre-processed near-infrared finger vein and high-resolution iris scan datasets. The attention mechanism is utilised in the network to identify region-of-interest (ROI) locations towards better performance based on classification. Multimodal feature comparison is performed using a similarity measurement based on a deep learning system within a fusion algorithm. System performance is measured with precision, recall, F1-score, and equal error rate (EER). The model is designed to enable strong feature extraction using deep learning libraries such as TensorFlow and Keras for model training and testing. Experimental results show that the composite biometric system performs greatly better than unimodal systems with a success rate larger than 97% and an EER smaller than 1.2%
  Keywords: Finger Vein Recognition; Iris Pattern Recognition; Convolutional Neural Networks (CNN); Deep Learning; Biometric Authentication; Feature Extraction; Multimodal Biometrics; Pattern Matching.
  DOI: 10.1504/IJIMS.2027.10074610
   
  
• Zero-Day Intrusion Detection System: Training on Small Datasets with Advanced Data Augmentation and Ensemble Learning  
  by Mehaboob Mujawar, Aasheesh Raizada, Abdullah Gubbi 
  Abstract: This paper presents a novel approach to developing a Zero-Day Intrusion Detection System (IDS) capable of identifying new or unknown cyber-attacks without retraining. The research addresses the challenge of training intrusion detection models on limited-sized datasets by employing advanced data augmentation techniques and ensemble learning methods. The proposed system utilizes a combination of Isolation Forest, Autoencoder, and Random Forest models to enhance detection capabilities. The system uses adaptive threshold methods, reconstruction errors to detect anomalies, and multi-model voting. Our research shows considerable increases in detection performance, especially for DDoS and DoS attacks. The system detects Heartbleed (100% accuracy), DoS Slowhttptest (94.58% accuracy), and DoS GoldenEye (93.45% accuracy). The approach's effectiveness is supported by ROC-AUC scores of 88.39% and 80.49% on various test days. The article also presents an interactive user interface for real-time model training, testing, and performance analysis. This interface improves system usability and visualizes detection metrics. The research solves a major network security issue by detecting zero-day attacks with insufficient training data
  Keywords: Zero-Day Intrusion Detection; Small Datasets; Data Augmentation; Ensemble Learning; Cybersecurity; Machine Learning; Anomaly Detection; Performance Metrics; Unknown Attacks.
  DOI: 10.1504/IJIMS.2028.10075219
   
  
• Revolutionising Higher Education: Unveiling the Impact of Information and Communication Technology on Student Achievement in Comparison to Traditional Teaching Methods  
  by Kabita Kumari Chaturvedi, Charu Bisaria, Prabhat Kumar Dwivedi 
  Abstract: Background ICT integration in higher education has transformed teaching methods, offering innovative ways to enhance student learning, performance, and skill acquisition. Objective This study assesses the impact of ICT-based teaching on student achievement and skills in Indian higher education institutions. Problem Statement Although ICT adoption is increasing, there is limited empirical evidence in India on its direct effects on student performance and skill development. Methodology A field-based survey was conducted among students and faculty in selected institutions using structured questionnaires to collect data on ICT use, academic achievement, and skill acquisition. Findings Results show that ICT-integrated teaching significantly improves student performance and enhances skill development compared to traditional methods. Conclusion Effective ICT integration leads to better learning outcomes. The study recommends policy support, institutional strategies, and further research to strengthen ICT-based teaching and promote innovation in higher education.
  Keywords: ICT; students; achievement; performance; teaching method; higher education system.
  DOI: 10.1504/IJIMS.2028.10076312
   
  
• Examine Consumer Intention to Adopt Driverless Delivery Vehicles: A Study Using PLS-SEM and NCA Approaches  
  by Rami Farhat, Qing Yang 
  Abstract: The rapid advancement of autonomous driving technology has led to the emergence of driverless delivery vehicles (DDVs). To promote widespread adoption, it is essential to address consumer concerns about safety, reliability, psychological factors, and the need for supportive policies that safeguard public safety and privacy. This study aims to understand and predict consumer adoption intentions for DDVs by extending the technology acceptance model (TAM) to include technological complexity and perceived trust. Data collected from 579 Lebanese respondents were analysed using partial least squares structural equation modelling (PLS-SEM) and necessary condition analysis (NCA). The results indicate that perceived ease of use (PEU), attitude, perceived trust, technological complexity, and perceived usefulness (PU) significantly influence consumer adoption intentions. Moreover, perceived trust and technological complexity indirectly affect adoption intentions through PU and PEU. The findings offer practical insights, recommending that business leaders collaborate with Lebanese regulatory authorities to ensure that DDV technologies meet local standards and regulations, thereby facilitating smoother integration into the market.
  Keywords: Driverless delivery vehicles; Perceived trust; Technological complexity; Necessary condition analysis; structural equation modeling.
  DOI: 10.1504/IJIMS.2028.10076524
   
  
• EXPLORING THE POTENTIAL OF ECO-FRIENDLY CUTTING FLUIDS IN CNC MACHINING: A MACHINE LEARNING-BASED APPROACH  
  by Srinivas Paleti, Padmini R, B. Srinivasa Prasad 
  Abstract: This work presents the potential of green-cutting fluids obtained by blending nano titanium di oxide in natural base oils in CNC machining. The cutting fluids are prepared using vegetable oils and nano additive at varying percentage inclusions (0.3% to 1.2%). The parameters considered for performance analysis are viscosity, specific density, thermal conductivity, absorbance by way of thermo-physical, rheological, and spectral properties. After, testing and predicting the basic properties, the prepared cutting fluids are applied to machining zone while turning, performance was investigated by considering the temperatures. Surface plots confirmed the rise in thermal conductivity with an increase in nano additive percentage. Cutting temperatures are modeled using novel machine learning algorithms including SVM, ANN, GPR and comparative analysis of the results reveals that optimum GPR is the best of the three considered with 87% fit while for SVM and ANN it is 82% and 63%.
  Keywords: Eco-safe cutting fluids; machine learning algorithms; temperatures; GPR; SVM.
  DOI: 10.1504/IJIMS.2028.10076530
   
  
• Left Ventricle Extraction Using a DENSENET-U-Net Integrated Deep Learning Approach  
  by X. Ignatious Viola, R. Shoba Rani, G. Savitha, R. Surekha 
  Abstract: The estimation of clinical parameters like stroke volume and ejection fraction relies on precise left ventricle (LV) segmentation from CT images. Many people have been paying attention to and hoping for deep learning-based image segmentation algorithms lately. Image analysis is a good fit for convolutional neural networks (CNNs), a kind of deep learning. Because it may simplify images by dividing things to be analysed into foregrounds and those to be ignored into backgrounds, segmentation finds widespread usage in medical image analysis. The purpose of this work is to provide a DenseNet-U-Net hybrid for left ventricle segmentation in computed tomography (CT). The segmentation result was improved by combining the DenseNet with the U-Net model with the ADAM optimiser. To begin, a clustering technique that utilises the available information about the spatial relationships between slices to crop the ROI has been proposed as a means to reduce memory consumption, processing time, and the class imbalance between the background and the target. We found that our suggested hybrid system outperformed the competition in terms of accuracy and performance in our experiments.
  Keywords: Cardiovascular diseases (CVDs); Left ventricle (LV); Image Segmentation; CNN; Deep Learning; U-Net; DenseNet.
  DOI: 10.1504/IJIMS.2028.10076630
   
  
• Real-Time Detection of Multilingual Audio Distress Signals Using Compact CNN and MFCC Features in Noisy Environments  
  by Shruthi G, Jinka Madhurilatha, Deepak D. J, Sowmya B. K, Nayana Rani S 
  Abstract: In order to find early warning signs of cardiac-related emergencies in acoustically demanding circumstances, this study explores a deep learning framework for the identification of human distress vocalizations To guarantee acoustic variety, a bespoke dataset was created that included recordings of background noise and common distress noises in Kannada and English Mel-Frequency Cepstral Coefficients (MFCCs) were used to convert audio inputs into spectral features that are perceptually significant To improve generalisation, data augmentation methods such as spectrogram masking, pitch shifting, temporal stretching, and additive noise were used To learn discriminative patterns from MFCC representations, a compact convolutional neural network (CNN) architecture was created, utilising max pooling, dropout, and batch normalisation to guarantee training regularization and stability The model's potential as a reliable tool for real-time distress identification was validated by its remarkably high accuracy (92%) and excellent F1-score (0 91) in identifying distress vocalisations This feature establishes the model as a practical means of improving security and safety by providing accurate, timely detection of important health events in a variety of settings.
  Keywords: Audio Distress Detection; Multilingual Speech Recognition; Mel-Frequency Cepstral Coefficients (MFCC) ,Convolutional Neural Network (CNN),Real-Time Audio Classification.
  DOI: 10.1504/IJIMS.2028.10076632
   
  
• AI-Enhanced Wireless Sensor Network for Proactive Structural Health Monitoring in Large-Scale Infrastructure  
  by Qinzhe Zhang 
  Abstract: Rising structural failures require structural health monitoring (SHM) systems to increase infrastructure life and resilience. A deep learning-based automatic surface change recognition method for structural health monitoring (SHM) systems using the Kaggle/Mendeley Surface Crack Detection Dataset is presented in this paper. The 40,000 227
  Keywords: Surface Crack Detection; MobileNetv2; Structural Health Monitoring (SHM); Deep Learning; Wireless Sensor Networks (WSNs).
  DOI: 10.1504/IJIMS.2028.10076794
   
  
• A Hybrid Huffman-CNN-AES Deep Learning Framework for Secure Medical Image Steganography: H-CADStego  
  by RAMAPRIYA B, Y. Kalpana 
  Abstract: As medical data becomes digital, protecting sensitive medical images is crucial. The hybrid framework H-CADStego uses Huffman encoding, convolutional neural networks (CNNs), and the advanced encryption standard (AES) to improve medical image steganography security and efficacy. The framework compresses the secret image using Huffman encoding to reduce redundancy and protects it with AES encryption. Train a deep learning model to embed the compressed and encrypted secret image into a cover image to create a visually identical steganography image. Use the same model to extract the secret image from the steganography image for precise reconstruction with minimal quality loss. PSNR, MSE, BER, SSIM, and correlation coefficients are used to evaluate the framework. High PSNR and SSIM scores confirm the steganography images imperceptibility, while low BER and MSE assure minimal image distortion. Compression, noise addition, and cropping attacks were not able to defeat the system. H-CADStego secures medical images with encryption, compression, and deep learning-based steganography. This method meets EHR, telemedicine, and secure data archiving privacy needs.
  Keywords: Steganograpy; Huffman encoding; peak signal-to-noise ratio; mean squared error; structural similarity index measure.
  DOI: 10.1504/IJIMS.2028.10076831
   
  
• A Hybrid PSO - SVM Algorithm for Enhanced Nodule Classification in CT Scans: Advancing Lung Cancer Diagnosis  
  by Shalini S, P.S.Eliahim Jeevaraj 
  Abstract: Machine learning-based lung cancer detection and classification may enhance patient outcomes and early diagnosis, attracting attention. This study compares four classification models: SVM, DNN, RF, and PSO-SVM (98.2%), surpassing DNN (97.7%), SVM (96.5%), and RF (95.8%). PSO-SVM (98.85%), DNN (98.08%), SVM (97.29%), and RF (96.45%) remain successful after sensitivity analysis. In addition, PSO-SVM had the best specificity (97.55%) and lowest false-positive rate (3.2%), ensuring classification dependability. The proposed PSO-SVM method has the best diagnostic performance and takes 72.34 seconds to compute. The wavelet transform has the highest accuracy of 97.00% and the lowest false-positive rate of 4.80% among feature extraction algorithms. Overall, the PSO-SVM model diagnoses lung cancer best. The system scales datasets and varies sensitivity to imaging noise in nodule density and shape to improve smart diagnostic assistance by optimising SVM training to tune the classifier to medical image nuances. Methodological and practical novelty add to the papers computational intelligence lung cancer diagnosis contribution.
  Keywords: Lung Cancer; Machine Learning; Particle Swarm Optimization (PSO); Deep Neural Networks; Random Forest; Sensitivity; Specificity; False-Positive Rate.
  DOI: 10.1504/IJIMS.2028.10076884
   
  
• Generative AI as a Catalyst for Intelligent Information Service Empowerment in University Libraries  
  by Li Zhao 
  Abstract: The rapid expansion of generative AI technology has created new opportunities and challenges for university libraries' information services. Generative artificial intelligence can manage knowledge, intelligently retrieve material, and provide targeted suggestions using deep learning and natural language processing, helping library services evolve. In this work, generative artificial intelligence is used to improve library information services such intelligent interactive services, data administration, autonomous knowledge organisation tool updates, and literary search tool optimisation. These initiatives can assist university libraries meet customers' wide and complex information needs, boosting library sector growth and service quality and efficiency. This study examines how Generative Artificial Intelligence (GAI) can transform university library intelligent information services. Librarians from five core institutions, professors, and students provided data for a mixed methods study. Comparing digital literacy, GAI use, training, and user happiness. The study uses advanced analytical methods like time series forecasting, logistic regression, multiple regression, and ANOVA. GAI frequency, digital literacy, and prior training strongly affect client happiness and service performance. Moderation research shows that training moderates the relationship between user happiness and GAI use. K-means clustering found three user typologies; ARIMA-based projections show AI-related questions rising exponentially.
  Keywords: Generative Artificial Intelligence; Information Service Management; University Libraries; Digital literacy.
  DOI: 10.1504/IJIMS.2028.10076934
   
  
• Smart Integration and Optimisation of Electronic-Electrical Architectures for Next-Generation Energy and Intelligent Connected Vehicles  
  by Haoxuan Jin, Guohong Yu 
  Abstract: Intelligent connected vehicles (ICVs) and new energy vehicles (NEVs) are increasing rapidly, necessitating reliable problem detection systems for predictive maintenance. A defect categorisation and predictive maintenance system will be created by deep and machine learning. With 11,000 sensor data points for voltage, current, motor speed, temperature, vibration, ambient temperature, and humidity, defects and maintenance needs were discovered. Random forest, gradient boosting, XGBoost, and LightGBM feature selection indicate that voltage, which accounts for 40% of the models decision-making, is the most important factor in fault classification. Gradient boosting, LightGBM, and bagging + boosting hybrid improved SVM classification from 61% to 93%. Ensemble learning excelled. Confusion matrices showed that LightGBM and XGBoost performed best at classifying inverter and battery issues. K-fold cross-validation indicated that LightGBM and XGBoost achieved the highest accuracy, over 94%, thereby enhancing model resilience. In regression analysis for predictive maintenance, random forest and LightGBM produced R2 values of 0.9178 and 0.9126, respectively. Time-series forecasting hybrid ARIMA-LSTM models identified linear and nonlinear fault patterns better than standalone models. High hyperparameters improved model efficiency and fault prediction. These findings demonstrate ML and DLs NEV and ICV diagnostic and predictive maintenance capabilities. Design minimises unexpected failures, maximises maintenance schedules, and detects issues in real-time.
  Keywords: Intelligent Connected Vehicles; New Energy Vehicles; Fault Detection; Machine Learning; Predictive Maintenance; Deep Learning and Optimisation.
  DOI: 10.1504/IJIMS.2028.10076940
   
  
• Enhancing Patient Data Security in Hospital Information Systems (HIS) through Blockchain Technology  
  by Bendi Venkata Ramana, W.Ancy Breen, Gayatri Mirajkar, V.P. Murugan, Nidal Al Said, T.R. Suchithra 
  Abstract: Health digitization has revolutionized the patient's life, but has posed unforeseen challenges in patient information security. Hospital Information Systems (HIS) now face cyber threats, leading to data breaches undermining patient confidentiality and trust. Blockchain technology, based on a decentralized, tamper-evident ledger, has immense potential to address this issue. This paper examines the application of blockchain technology in the healthcare industry to enhance the security of patient information. We explore current vulnerabilities in HIS, examine current literature on healthcare blockchain implementations, and propose a blockchain framework specially tailored for HIS. We determine, through simulation and case studies, the relevance of applying blockchain to prevent data breaches and unauthorized access. The outcomes indicate that unauthorized access cases and data breaches are significantly minimized. We also address the challenges of adopting blockchain, such as scalability and interoperability issues, and provide future directions to address these challenges. The outcomes illustrate the possibility of applying
  Keywords: Blockchain Technology; Patient Data Security; Hospital Information Systems; Data Integrity; Healthcare Cybersecurity.
  DOI: 10.1504/IJIMS.2028.10077165
   
  
• Dental Image Analysis for Anomaly Detection and Classification with CBAM Efficient NET B0  
  by S. Srividhya Santhi, R. Shoba Rani, S.Sengamala Barani, K. Satyanarayana 
  Abstract: A variety of oral health issues may manifest as dental abnormalities (DA), which can lead to pain, infection, and even tooth loss. The majority of people have dental caries at some point in their lives. A person's quality of life may take a nosedive when they experience excruciating pain or illness due to dental caries. Early detection and treatment of dental caries can be achieved using machine learning models. Nevertheless, doctors often express dissatisfaction with the model's outcomes since they are not easily explicable. Dental disorders must be diagnosed quickly and accurately to prevent complications. Over time, untreated oral infections can cause tooth loss. Dentists depend on dental X-rays to diagnose oral diseases. If teeth cavities can be automatically detected, treatment may be faster and cheaper. Many studies have attempted to develop robust deep learning models to identify DA in photos. This study employed a hybrid EfficientNet B0-CBAM model to classify dental pictures for dental caries detection. Dental image segmentation using an upgraded MLP-UNet architecture detects anomalies. The results show that our approach accurately diagnoses dental defects, such as caries, better than others.
  Keywords: Dental Anamoly detection; segmentation; classification; MLP; UNet; EfficientNet B0; CBAM; NAG; CNN.
  DOI: 10.1504/IJIMS.2028.10077449
   
  
• Wear Properties Characterization of manufactured Mg-HAP Matrix Composites for the Purpose of Medical Biotechnology  
  by Neeraj Kumar, R.A.J. Kumar Duhan 
  Abstract: A novel class of biocompatible implant materials called magnesium matrix composites will unavoidably experience concurrent deterioration and rust throughout the human body. The purpose of this work is to examine the wear behavior of hydroxyapatite in a magnesium matrix composite. Magnesium corrodes too fast in the environment of the body; matrix composites provide a way to modify the material's wear characteristics. The Mg-HAP MMCs were produced in the present study via stir casting. The use of HAP powder is one kind of reinforcing material. The required form samples for metallurgists to evaluate were created from the cast composites. The magnesium-HAP composite's wear characteristics were examined using a Pin on Disc Tribometer for a dry sliding wear test. Characterizing wear testing was done at the last phase. Both rate of wear and rate of specific wear greatly boosted by the addition of HAP powder. Lower frictional coefficient and wear resistance achieved.
  Keywords: Biocompatible; Magnesium Hydroxyapatite; Metal Composites; Stir Casting; Wear properties; Medical Biotechnology.
  DOI: 10.1504/IJIMS.2027.10077637
   
  
• Security in Microkernel-Based Operating Systems: A Comprehensive Review  
  by Ahlam Abdulbaset Ali Saif, Ali Haider Shamsan, Nail Adeeb Ali Abdu 
  Abstract: Microkernel-based operating systems present a robust alternative to monolithic kernels, particularly in security-sensitive environments. Their modular architecture relocates non-essential services to user space, thereby reducing the trusted computing base (TCB), minimising attack surfaces, and enhancing fault isolation. This paper offers a comprehensive analysis of the security benefits and challenges of microkernels, including mitigation of privilege escalation, side-channel attacks, and formal verification complexities. We also highlight recent advancements such as time protection and secure IPC and explore future directions involving AI, blockchain, and quantum computing integration. The findings underscore microkernels potential in building resilient and secure next-generation operating systems
  Keywords: Operating System TBC; IPC; Microkernel; Performance; Security.
  DOI: 10.1504/IJIMS.2028.10079025