This study assessed ICTMT awareness among Indian and Hungarian university students on different continents. He analyzed student attitudes toward ICTMT using differential, inferential, and predictive methods. MLR approaches predicted high Hungarian university student opinions. According to Pearson correlation, technology benefits favorably correlated with student attitude and usability. Exploratory Factor Analysis (EFA) suggested important MLR model elements that predicted student evaluations using educational benefit and usability factors. Thus, Hungarian students believe technology usability and benefits have been greatly affected. Based on ICTMT technology, Indian students' attitude was identified with the LR method. Their attitude and ICTMT advantages correlated positively according to the Pearson Correlation. EFA-PCA recommended all features. The Mann-Whitney U test then compared Indian and Hungarian students' technology use and advantages opinions. The T-test also promotes technology's national benefits. Technology use among Indian and Hungarian students was correlated through Correspondence Analysis (CA). It found a considerable technology benefit for Hungarian pupils but not Indian students. The Chi2, Fisher's Exact (FE), and Cramer's v (CV) tests assessed their viewpoint. The Mann-Whitney U test showed that Hungarian and Indian pupils' opinions differed. Then, Kruskal-Wallis H, Welch's t, and Mann-Whitney U tests compared opinions on technological progress and access. These statistical tests show that Indian and Hungarian students' perspectives differ. The optimistic machine learning models SVM, MLP, and PCA predicted students' nations based on their technology opinions. Another machine learning predictive model identifies ICTMT students' gender and geography. Students, teachers, parents, institutions, and government may benefit from the research. Students in both nations can learn about education technology developments. Management, technical coordinator, and teachers can assess and train pupils' technology awareness.

He developed Student Demographic Identification (SDI) to accurately identify a student's demographic attributes (age, course).  SDI was compared to Random Forest (RF), Logistic Regression (LR), and Radial Support Vector Machine (R–SVM). The suggested technique improves classifier accuracy, F1-score, precision, recall, and Matthews Correlation Coefficient (MCC). He suggested important features to identify students' age, course, and gender. SDI correctly identified the student's age group and course 96% and 97%, respectively. Gradient Boosting (GB) improves LR, R-SVM, and RF gender prediction. RF and GB also identified student genders with 98% accuracy. All three classifiers also correctly recognized the student's area and institution with 99% accuracy. Our SDI algorithm may predict student demographics in real-time surveys.

Using machine learning and statistics, this study examined innovative elements affecting informatics students' satisfaction and happiness. The Random Forest (RF) method accurately predicted student happiness at 88%. Students loved hybrid learning and thought it was better for theory than laboratories. Student's t–test showed that age and course affected hybrid learning safety online (p < 0.05). Students in bachelor's programs (B.Sc.) under 20 were more confident than those in master's programs (M.Sc.) over 21 that hybrid learning is safe during the epidemic. Student satisfaction affects online education safety concerns and prospects. Time and expense utilization with management, University activities and assistance, and hard group work significantly affected student happiness. Future hybrid learning style, Appropriate for theory classrooms and laboratories, and Time and cost utilization with management were important predictors of student satisfaction with all machine learning methodologies. The article suggested hybrid learning challenges, pandemic support, and safety measures for higher education institutions.

A study conducted at Changhua Christian Hospital in China used serum biomarker samples to classify liver disease stages. The researchers used machine learning algorithms like Random Forest, Logistic Regression, XGBoost, and Support Vector Classifier to predict liver risk associated with conditions like Hepatitis, Autoimmune Hepatitis, Alcoholic Liver Disease, and NonAlcoholic Fatty Liver Disease. The study identified three distinct clusters using LSA, with the RF model achieving high accuracy of 0.94±0.06. Key features in the liver disease staging increment were identified, including GPT, Age at Diagnosis, Erythrocyte Sedimentation Rate, and C-reactive protein. The developed model provides valuable decision-making tools for clinicians, enabling early and targeted interventions in liver disease management.

Since the COVID-19 pandemic, teachers and students have started using online and hybrid learning in education. There might be several obstacles to adopting hybrid learning in theory classes or lab practice sessions. Based on student opinions, deciding what is appropriate for theoretical class and lab practice is challenging. We employed machine learning approaches to forecast the hybrid learning mode for theory classes and lab practices. We introduce a framework that utilizes machine learning to automate the identification of hybrid learning for Theory Class and Lab practice (TCLPI). Four machine learning models form the foundation of this framework: Random Forest (RDT), Support Vector Machine (SVN), Logistic Regression (LGR), and Extreme Gradient Boosting (XBT). In the context of Theory Class Identification (TCI), the SVN achieves a maximum test accuracy of 0.93, whereas the LGR achieves a minimum accuracy of 0.90. On the other hand, the Lab Practice Identification (LPI), XBT, RDT, and SVN achieved a test accuracy of 0.80. The outcome of trained algorithms is assessed using the Shapley Additive Explanation (SHAP), an explainable Artificial intelligence (AI) approach. This research found that student-teacher interaction decreased during lab practice, which is crucial. Internet disconnections, a lack of support during technological malfunctions, and the likelihood of cheating in exams without monitoring are also issues. We also found that students were accepting of hybrid learning for theory classes. Each model’s intrinsic feature relevance and SHAP values helped prove this. Research shows that hybrid learning works more for theory classes; it is less needed for lab practice for students.

The security of 3-D images is an important research problem that has to be resolved. Due to the more complex structure of 3-D images, the encryption of these images is quite different from 1-D and 2-D images. A threestage novel image encryption technique based on chaotic maps is presented in this paper in which a 3-D image is firstly converted to a similar image format as that of 2-D images before encryption. The initial conditions of the chaotic system are generated by using the SHA-256 function on the coordinate matrix of the plaintext. Initially, a logistic map is utilized to scramble and add random points to the coordinate values of vertices of a 3-D image. The coordinate values are then confused and diffused in the second stage by using three sequences generated through the logistic-dynamic coupled logistic map lattice (LDCML) model. This stage also involves splitting of floating point data into integer and fractional parts. The integer part is diffused whereas the fractional part is scrambled during this process. In the third stage, the confusion is performed by using a tent map among the coordinate points. This process enhances the robustness, integrity, and confidentiality of 3-D images and ensures protection against unauthorized access. The proposed encryption procedure achieves values of correlation close to zero along x,y,z- directions, NPCR of 100%, UACI of 33.37%, and entropy value of 7.9993 which demonstrates its robust security. The time analysis shows that our technique improves efficiency and lowers computational costs by processing data in a lesser time. The security and statistical analysis concludes that the proposed encryption algorithm can withstand various conventional attacks.

Background and Objective: Early detection of eye diseases, especially cataracts, is essential for preventing vision impairment. Accurate and cost-effective cataract diagnosis often requires advanced methods. This study proposes novel deep learning models that integrate global and local attention mechanisms into MobileNet and InceptionV3 architectures to improve cataract detection from fundus images. Methods: Two deep learning models, Global–Local Attention Augmented MobileNet (GLAAM) and Global– Local Attention Augmented InceptionV3 (GLAAI), were developed to enhance the analysis of fundus images. The models incorporate a combined attention mechanism to effectively capture deteriorated regions in retinal images. Data augmentation techniques were employed to prevent overfitting during training and testing on two cataract datasets. Additionally, Grad-CAM visualizations were used to increase interpretability by highlighting key regions influencing predictions. Results: The GLAAM model achieved a balanced accuracy of 97.08%, an average precision of 97.11%, and an F1-score of 97.12% on the retinal dataset. Grad-CAM visualizations confirmed the models’ ability to identify crucial cataract-related regions in fundus images. Conclusion: This study demonstrates a significant advancement in cataract diagnosis using deep learning, with GLAAM and GLAAI models exhibiting strong diagnostic performance. These models have the potential to enhance diagnostic tools and improve patient care by offering a cost-effective and accurate solution for cataract detection, suitable for integration into clinical settings.

Image ciphering techniques usually transform a given plain image data into a cipher image data resembling noise, serving as an indicator of the presence of secret image data. However, the transmission of such noise-like images could draw attention, thereby attracting the attackers and may face several possible attacks. This paper presents an approach for generating a visually meaningful image encryption (VMIE) scheme that combines three layers of security protection: encryption, digital signature, and steganography. The present scheme is dedicated to achieving a balanced performance in robustness, security and operational efficiency. First, the original image is partially encrypted by using the RSA cryptosystem and modified Hénon map (MHM). In the second stage, a digital signature is generated for the partially encrypted image by employing a hash function and the RSA cryptosystem. The obtained digital signature is appended to the partially encrypted image produced after implementing the zigzag confusion in the above partially encrypted image. Further, to achieve better confusion and diffusion, the partially encrypted image containing a digital signature undergoes through the application of 3D Arnold cat map (ARno times), to produce the secret encrypted image (Sr5). To ensure the security and robustness of the proposed technique against various classical attacks, the hash value obtained from the SHA-256 hash function and carrier images is utilized to generate the initial conditions Mh10 and Mh20 for modified Hénon map, and initial position Zip = (zrow, zwcol) for zigzag confusion. In the proposed algorithm, the digital signature is utilized for both purposes to verify the sender’s authenticity and to enhance the encryption quality. The carrier image undergoes lifting wavelet transformation, and its high-frequency components are utilized in the embedding process through a permuted pattern of MHM, resulting in a visually meaningful encrypted image. The proposed scheme achieves efficient visual encryption with minimal distortion and ensures lossless image quality upon decryption (infinite PSNR), balancing high level of security along with a good computational efficiency.

Background & Objective

Liver disease remains a significant global health burden, often progressing silently until advanced stages such as cirrhosis or hepatic failure. Early detection is essential but remains hindered by the limitations of conventional diagnostics. This study presents LivXAI-Net, an explainable artificial intelligence (XAI) framework integrated with Internet of Things (IoT) biosensors, designed and evaluated in a simulated real-time setting using a historical dataset.

Methods:

LivXAI-Net simulates continuous data acquisition from wearable biosensors — including sweat, platelet, and prothrombin sensors — and processes this data using machine learning (ML) models trained on the Mayo Clinic primary biliary cirrhosis (PBC) dataset (n=

424, 1974–1984). Random Forest (RF) and XGBoost(XGB) classifiers were deployed with SHAP and Permutation Feature Importance (PFI) to enhance interpretability. A mobile application, Hepatic Health Tracker, delivers real-time risk predictions, supported by a secure data pipeline using TLS 1.3 and AES-256 encryption.

Results:

RF and XGB achieved accuracies of 84% and 82% respectively under 20-fold cross-validation. Key biomarkers — albumin, cholesterol, and triglycerides — were consistently identified by SHAP as influential in classification. The system achieved a total latency of 0.85 s in a simulated 5G environment, supporting near-instantaneous alert delivery via the mobile interface.

Conclusion:

LivXAI-Net combines interpretable ML with real-time biosensor data to enable proactive liver disease management. While currently validated using historical data in a simulated environment, future work will involve deployment with live sensor input and clinical trials to validate utility and generalizability in real-world settings.

Recent technological advancements have facilitated the widespread integration of robots across various sectors, including healthcare, defense, and industries. The Internet of Things (IoT) based robotic devices, equipped with advanced sensors and imaging capabilities, are crucial for various tasks such as environmental monitoring, autonomous navigation, and real-time surveillance. For real-time surveillance, robotic devices contribute to effective, safe, and reliable communication. However, the communication, in real-time environment, faces several challenges such as security and resource optimization. Due to limited computational resources, IoT-based systems need lightweight security solutions for communication efficiency and secure data handling. To enhance communication efficiency with secure data handling, this study introduces a Lightweight Surveillance Framework for Robotic Systems (LSFRS). The LSFRS optimizes data transmission by leveraging the Structural Similarity Index to eliminate redundancy and improving resource utilization. For secure communication, lightweight encryption methods: Tiny Encryption Algorithm (TEA), eXtended TEA (XTEA), Caesar Cipher, and Exclusive or (XOR), are employed. The proposed framework is evaluated for encryption performance, computational efficiency, and suitability for real-time surveillance. Comparative analysis demonstrates that the proposed framework is competitive against existing techniques.

Neuro-imaging data is necessary for timely and correct diagnosis of the disease Alzheimer’s (AD). Early clinical intervention is challenging due to the high-dimensional nature of the images, making it computationally difficult to work on representations, sparse annotated datasets and inter-subject variance. Deep convolutional VGG, ResNet, DenseNet, MobileNet and other neural networks (CNNs) have shown good performance. Diagnostic performance, but they are based on a high number of parameters and require highly trained personnel. Data can often times restrict resilience to data-skimy clinical settings. To address these challenges, the current paper presents a hybrid Quantum Machine Learning (QML) architecture for Alzheimer’s disease Detection and classification based on a combination of classical CNN-based feature extraction and Variational quantum classifier. In the suggested method, diseased, relevant, and compact features are extracted using pre-trained CNN backbones. Data from structural MRI scans is later translated into quantum states and computed. With parameterised variational quantum circuits optimised with a hybrid quantum-classical learning loop. The proposed method has been experimentally evaluated on benchmark datasets of MRI scans of Alzheimer’s patients. The hybrid CNNQML model has a mean classification accuracy of about 93% which is similar to deep CNN models, including ResNet and DenseNet (94%), and much more. Beating lightweight models like MobileNet and shallow CNNs by six percentage points. It is worth noting that the hybrid QML model performs better in conditions with limited training data. Performance degradation is observed in accuracy and AUC for all CNN variants, with up to 5% lower performance to deep CNNs. Cross-validation, confidence interval, paired significance test, and statistical validation. Effect size analysis proves the fact that the hybrid model produces statistically significant performance, as well as cases of clinically meaningful improvements over lightweight CNN baselines. Similar to the state-of-the-art deep networks with significantly reduced trainable parameters in the classification stage. These findings show that hybrid QML is a data-efficient and powerful substitute in the diagnosis of Alzheimer’s. On the whole, this is a work that sets a practical way to integrate quantum machine learning into clinical neuro-imaging pipelines and hybrid settings. QML has supplementary deep-learning enhancement to classics in next-generation Alzheimer’s Decision-support systems.

AI tools such as ChatGPT are used in the higher education increasingly yet the limited empirical evidence explains how the students incorporate these tools into their academic work. This study examines how students’ perceptions and demographic characteristics relate to their intended use of ChatGPT academically. The study focuses on understanding the educational meaning rather than maximizing the algorithmic performance. The survey data were collected from 166 university students for this purpose. It is analyzed to identify the key learning related constructs associated with ChatGPT use. An exploratory machine learning (ML) framework was applied to examine the relationships among these constructs while maintaining the interpretability.

The analysis suggests that learning support, accessibility and academic engagement are the important factors that were associated with students’ intentions to use ChatGPT in academic contexts. These elements indicate that students often view AI tools as resources for clarifying the concepts, supporting coursework and enhancing the learning efficiency. At the same time, the concerns related to reliability, accuracy, and academic integrity influence how students evaluate the appropriate use of such tools. Overall, the findings suggest that ChatGPT adoption in higher education is shaped largely by the pedagogical and contextual conditions rather than technological performance alone. The study provides the exploratory insights that may inform the instructional design, AI literacy initiatives, and the responsible integration of generative AI (GAI) within the higher education environments.

Skin diseases frequently cause mental and physical distress and are a major global health concern. Because early detection is crucial to successful treatment, accurate diagnosis is a challenge for dermatologists as well. Diagnostic accuracy could be significantly enhanced using methods like machine learning (ML) and deep learning (DL). However, substantial datasets are required for these models to make accurate predictions. The healthcare providers frequently encounter data shortages, and privacy regulations restrict data sharing. A privacy-preserving federated transfer learning for diagnosing skin diseases which incorporates four key strategies to enhance effectiveness. The transfer learning is used to train a model with dense neural network (DNN) for skin disease detection. The feature extraction is performed using pre-trained architectures and DNN is used for classification. The federated learning (FL) replaces the transfer learning to train the model across distributed nodes with the DNN used for disease detection. The FL is combined with transfer learning to build a cohesive ecosystem where data privacy is maintained. The model performance was validated on both IID and non-IID databases, with the proposed feature extraction with federated learning model achieving cross validation accuracy of 99.528% and 99.689% for IID and non-IID databases, respectively. Results indicate that feature extraction with FL model can produce efficient, lightweight models—well-suited for resource-constrained devices—while ensemble learning enhances edge device performance, offering a powerful and privacy-preserving solution for skin disease diagnosis in modern healthcare.

Background and Objectives:

AI enabled recruitment systems suffered with the challenges like bias, transparency, and fairness in student placement processes. Traditional manual hiring methods are always found slow and subjective, leaves doubt of unintentional discrimination. This study proposes a data-driven recruitment framework that combines Natural Language Processing (NLP), Machine Learning (ML), and Explainable AI (XAI) to enhance résumé screening, predict placement outcomes, and assess fairness. It investigate the four key questions here (1) the extent to which NLP models improve the résumé screening accuracy; (2) the effectiveness of fairness oriented ML methods in reducing demographic bias; (3) the contribution of XAI to transparency and interpretability; and (4) also the role of Aspect-Based Sentiment Analysis (ABSA) in deriving the actionable insights from recruiter student feedback.

Methods:

The study uses 731 anonymized student records from the Computer Science and Engineering (CSE) department of a private university in Punjab, India, collected between the academic years 2020 and 2023. The proposed framework incorporate NLP techniques such as TF-IDF, Word2Vec, and BERT for automated résumé screening. Several ML classifiers, including Logistic Regression (LR), Support Vector Classifier (SVC), Random Forest (RF), and Extreme Gradient Boosting (XGB), are used for placement prediction. Fairness is evaluated through the Demographic Parity Difference (DPD), while Exponentiated Gradient Reduction (EGR) mitigates the demographic imbalances. Additionally, A fine tuned BERT based ABSA model analyzes the recruiter feedback to pinpoint area for improvement in the placement process.

Results:

The Empirical analyses validated all the four proposed hypotheses. The résumé-screening model achieved 98.15% accuracy, while placement prediction reached up to 81%. Applying EGR substantially improved the demographic parity. SHapley Additive exPlanations (SHAP) based explainability improved the interpretability, and the ABSA module achieved 97% accuracy, uncovering the valuable insights related to communication, logistics, and program design.

Conclusion:

The validated framework demonstrates that combining the fairness aware learning models with XAI significantly improves accuracy, equity, and even transparency. Future research will expand the dataset diversity, explore the additional fairness metrics, and will incorporate real-time recruiter feedback to develop adaptive, ethical, and human-centered hiring systems.

The rise of 3-D data dependency across various fields, from medical imaging to virtual reality, requires robust security to prevent leakage or unauthorized access and to keep sensitive information secure from tampering. As 3-D models consist of floating-point numbers arranged in a specialized format, their protection necessitates a reliable encryption technique to safeguard sensitive data across various domains. In this paper, a symmetric encryption technique for 3-D models based on the quantum logistic map, Arnold cat map, improved logistic map, and DNA encoding is presented. Initially, the hash value derived from the SHA-256 function, combined with the plaintext data, is used as a symmetric key to establish the initial conditions for the chaotic system. Following this, the improved logistic map (ILM) is applied to introduce confusion among the coordinate values of plaintext data. Subsequently, the partially encrypted data is segmented into integer and fractional components. The integer matrix undergoes scrambling and diffusion through the complete framework of the quantum logistic map, Arnold cat map, DNA diffusion algorithm, and DNA complementary rule, enhancing the complexity of the encryption process. On the other hand, the fractional part is manipulated using the quantum logistic map. Finally, the transformed integer and fractional parts are systematically integrated to obtain the final encrypted data. The proposed algorithm undergoes rigorous testing on a variety of 3-D models, validating the effectiveness of the proposed encryption scheme by offering a high entropy value close to 8, 100% NPCR and 33.37% UACI values, near-zero correlation, strong resistance against differential attack, maintained computational efficiency, and high key sensitivity. Furthermore, encryption time varies linearly with the size of input 3-D model, demonstrating the applicability of the proposed algorithm for large-scale 3-D datasets.

Optimization search algorithms play a crucial role in solving complex constrained engineering and Multi Layer Perceptron (MLP) classification problems, often requiring efficient and robust search strategies. This study introduces the Chaotic Zebra Optimization Algorithm (CZOA), an enriched version of the classical Zebra Optimization Algorithm (ZOA), incorporating chaotic maps to enhance search efficiency and convergence speed. The CZOA variants (CZOA1, CZOA2, CZOA3, and CZOA4) are evaluated using 23 standard benchmark functions across multiple dimensions (10D, 30D, 50D, and 100D), the CEC 2022 benchmark test suite, and three constrained engineering optimal design problems (three-bar truss, pressure vessel, and multi-disk clutch optimization). Furthermore, CZOA's effectiveness is evaluated on four MLP classification datasets (Balloon, Iris, Breast Cancer, and XOR dataset). Simulation results reveal that CZOA1 consistently outperforms other variants and classical ZOA, achieving the lowest fitness values, lowered standard deviations, and enriched convergence rates. Specifically, in F1 (10D), CZOA1 reduces computation time from 4559.9 s (ZOA) to just 0.1563 s, achieving faster speed up. In CEC 2022 functions test suit, CZOA1 delivers the best performance with an average fitness of 333.86 and a standard deviation of 0.69, surpassing state-of-the-art metaheuristics such as GWO, WOA, and TLBO. Engineering optimal design results confirm CZOA1's superior fitness values, with 263.8965 in the three-bar truss problem and 6849.485 in pressure vessel optimization, outperforming classical methods like GWO, DE, and PSO. For MLP classification tasks, CZOA1 achieves a 99.00% classification rate in breast cancer detection, highlighting its superior ability in handling high-dimensional search spaces. Comparative evaluations demonstrate that chaotic maps significantly enhance exploration–exploitation balance, reduce computational search complexity, and improve solution accuracy. The findings establish CZOA as a promising optimization technique for real-world engineering and classification challenges.