Authors - Nguyen Xuan Ha Giang, Lam Thanh-Toan, Nguyen Thai-Nghe Abstract - This study introduces a novel dual-branch Deep Matrix Factorization (DeepMF) framework enhanced by NLP techniques for predicting student performance in the Intelligent Tutoring Systems. Building on previous research, the proposed approach adopts a fundamentally different modeling strategy by transforming discrete educational features-such as learner ID, exercise, question, skill group, session start time, and repeated attempts-into structured sentences that capture both temporal and sequential information. These inputs are processed through two complementary branches. The first branch employs pre-trained GloVe embeddings, followed by a self-attention layer that captures intra-sequence dependencies before passing the representations into a DeepMF module. The second branch leverages a BERT-based model to extract contextualized language features. To address the issue of class imbalance, Focal Loss is applied during training on both the KDDCup 2010 and Assistment 2017 datasets. Experimental results demonstrate substantial improvements in prediction accuracy: RMSE is reduced from 0.418 to 0.167 (a 60.1% reduction) on KDDCup 2010, and from 0.472 to 0.186 on Assistment 2017 (representing a 60.6% relative improvement). These findings confirm the effectiveness of integrating contextual, sequential, and temporal modeling with DeepMF in educational data mining.
Authors - Hariprasad Rai M, Shankar Lingam. M Abstract - The COVID-19 pandemic has had profound and multifaceted impacts on global economies, with gender disparities becoming increasingly apparent. This paper explores the gendered effects of India's policy measures in response to the pandemic, focusing on the socio-economic challenges faced by women. It assesses the policy interventions implemented by the Indian government to mitigate these impacts, such as cash transfers, healthcare support, and social security measures. Drawing from data collected through the COVID-19 Global Gender Response Tracker, the study highlights the gendered dimensions of these policies and their effectiveness in addressing the unique needs of women during the crisis. The analysis underscores the need for gender-sensitive economic and social policies to ensure equitable recovery and resilience for women in future crises. By critically examining the intersection of gender, policy, and pandemic response, this paper contributes to a deeper understanding of the importance of gender-inclusive strategies in man-aging global health emergencies.
Authors - Md Mahbub Alam, Sabrina Sultana Prithul, Md. Sajjad Hossain Abstract - Understanding and predicting customer behavior is essential for sustaining growth and profitability in the software market. This study proposes an AI-driven framework for analyzing customer adoption patterns and churn risk through the integration of clustering and classification techniques. Using real-world behavioral data, the research first applies K-Means clustering to segment users into distinct behavioral groups. Four meaningful segments were identified, each characterized by varying engagement levels, contract types, and churn tendencies. Subsequently, supervised machine learning models—Logistic Regression, Random Forest, and XGBoost—were employed to predict churn. Among these, XGBoost achieved the highest performance, with an accuracy of 87.1% and an ROC-AUC score of 0.91. Feature importance analysis highlighted tenure, contract type, and monthly charges as critical churn predictors. The findings offer practical insights into personalized retention strategies and pricing interventions, emphasizing the utility of AI in customer analytics. This work contributes to the field by bridging behavioral segmentation with predictive modeling, providing a scalable and interpretable approach to managing customer lifecycle challenges in software-based services.
Authors - S. Rajaprakash, G. Sujatha, S. Kavitha, Vivek Darsi, Mohith Sai Kurakalva, Srikar Reddy Palapati Abstract - The growth of mobile applications has reached extraordinary levels because they provide users better features and convenience for daily needs. Massive mobile application expansion resulted in fresh security issues creating an in-creased number of cyber threats throughout the market. The number of cyber risks increases in Germany since malicious applications both harm device integrity and illegally obtain user information to threaten individual privacy together with organizational security measures. The development of strong auto- mated security systems with locating and removal capabilities of threats remains an urgent matter. Cybersecurity systems at present base their protection on fixed rule systems alongside traditional machine learning approaches. The implemented security methods deliver protection however they demonstrate reduced performance during changes in cyber threat patterns. The system exposes sever-al entry points which hackers can leverage for attack purposes. An innovative cyber- security framework has been developed which incorporates the RF algorithm with PCA through this project to address existing limitations. The system uses mobile application metadata preprocessing to accomplish three tasks through PCA: elimination of unneeded data and retention of important features along with feature selection. This procedure simultaneously maintains data authenticity as well as enhances processing speed. Random Forest becomes operational to classify mobile applications between malicious and safe categories within the framework. RF algorithm and PCA form a flexible and scalable system which enables handling extensive datasets along with security adaptation towards the latest threats. The system has a user-friendly interface which enables users to enter mobile app metadata and get prompt predictions along with performance scores. Mobile app users benefit from predictions which help them decide about the safety of apps through confidence scoring. The system maintains strict vali- dation procedures across unidentified datasets which confirm its dependable and practical functionality during actual use. Evaluation of the model effectiveness depends on performance metrics that include accuracy together with precision, recall and F1-score. The system delivers complete performance information which allows end-users to receive practical insights and enable greater transparency.
Authors - Gaurav Kamble, Chetan Parlikar Abstract - This study explores the impact of the quality of healthcare services, their accessibility, and affordability on patient satisfaction. With patient satisfaction becoming a crucial determinant of healthcare quality, it is important to establish drivers of patient satisfaction in order to empower healthcare providers to improve service delivery. Specifically, the study seeks to establish the level to which healthcare service quality, accessibility of healthcare services, and their affordability independently influence patient satisfaction. A survey of 206 respondents was conducted, and data on respondents' perceptions and experiences of healthcare services were collected. The study analyses the interaction among the three drivers and patient satisfaction with a view to providing actionable in-sights into the improvement of healthcare outcomes. With the onset of greater emphasis on patient-centred care, this study provides timely and relevant evidence capable of guiding policy and strategy towards enhancing the quality, accessibility, and affordability of healthcare services.
Authors - Jay Prakash Thakur, Akshata Kishore Moharir Abstract - Multi Agent Systems (MAS) face a critical challenge to maintain trustworthiness alongside operational efficiency as they move from theoretical development to practical implementation in essential domains. Human-in-the-loop approaches create operational bottlenecks that limit scalability, but fully autonomous systems face the risk of catastrophic failures. The proposed framework in the paper introduces a new approach to minimal human oversight through strategic intervention points that use intelligent triggers that assess risk and quantify uncertainty, detect novelty, and analyze consensus. The framework achieves scalable oversight of complex MAS through optimized information flow and reduced cognitive load that maintains critical safety guarantees.This position paper conducts a theoretical analysis of the impact of system performance while presenting applications for autonomous transportation, critical infrastructure management, and financial systems.
Authors - Sasipong Kijsason, Sa-Aat Niwitpong, Suparat Niwitpong Abstract - Parameter variance is used to measure the dispersion of data or the deviation of individual data points from the mean. A high variance indicates that the data are widely spread around the mean, reflecting greater variability within the dataset. This study introduces four novel methods for constructing confidence intervals for the variance of the Zero-Inflated two-Parameter Rayleigh distribution. These include the percentile bootstrap, the bootstrap method with standard error, the standard method based on a large sample, and generalized confidence interval approaches. A simulation-based comparison was conducted using coverage probability and expected length as performance criteria. The findings indicate that the generalized confidence interval and standard method achieved coverage probabilities closest to the nominal confidence level. Among these methods, the generalized confidence interval demonstrated the highest efficiency. Additionally, the proposed methods were applied to real-world data on COVID-19 mortality rates in Malaysia during September 2021.
Authors - Eshwari V. Kadu, Sudhir Agarmore Abstract - An IEEE family of standards called Time-Sensitive Networking (TSN) builds upon Ethernet to facilitate determinism over communications with guaranteed low latency, low jitter, and high dependability. Providing synchronized, deterministic network operation on prevalent Ethernet equipment, TSN finds itself at the center of essential enabler support for emerging applications such as industrial automation, self-driving vehicles, and smart grid, all which necessitate deterministic, real-time sharing of information. Time-sensitive scheduling, traffic priority, redundancy, and perfect time synchronization are only a few of the most important aspects of TSN discussed in depth in this paper, which also follows the development of the technology from ordinary Ethernet and identifies the necessity for time-sensitive communication. Besides, we examine the challenges of deploying TSN, summarize real-world applications across various industries, and discuss emerging trends such as integrating TSN with wireless networks and edge computing. As part of ensuring TSN has the ability to meet the demands of future-time-critical systems, we identified open research avenues and directions as part of our research.
Authors - Thanaporn Phattanaviroj, Massoud Moslehpour, Princy Pappachan, Mosiur Rahaman, Jirapong Pomnoi, Rinruedee Pattaradej Abstract - The tourism industry has undergone significant transformations following the COVID-19 pandemic, particularly through the adoption of digital technologies such as virtual tours. This study investigates the motivational factors influencing individuals' intention to adopt virtual tours, using Protection Motivation Theory (PMT) as the theoretical framework. It identifies two key components: threat appraisal (perceived travel and health risks) and coping appraisal (enjoyment, perceived cost-effectiveness, and sustainability motivation). Data was collected through a designed questionnaire and analyzed by applying Partial Least Squares Structural Equation Modeling, with 122 respondents. The findings indicate that threat and coping appraisals have a significant influence on the intention to adopt virtual tours. The research provides insight into the evolving behavior of tourism consumers in a post-pandemic context. It highlights the potential of virtual tourism as a sustainable and attractive alternative to physical travel.
Authors - Anju Kamal, Rajiv Prasad Abstract - The emergence of technology driven gig economy and digital labor platforms has transformed the methods of sourcing, assessing, and remunerating work. This study examines how traditional credentials like educational degrees and platform-specific ratings like Top Rated and Top-Rated Plus affect freelancer earnings in digital labor marketplaces. We use regression analysis to compare the explanatory power of ratings given by the platform and educational levels to determine whether platform ratings are more strongly associated with hourly earnings than formal education credentials for 146 freelancers on a prominent digital labor platform. Based on signaling theory, we found that platform-assigned ratings that acts as signals predict hourly earnings better than formal education. The findings indicate that digital labor marketplaces value platform signals like innovative reputation systems. The results show the association between traditional educational credentials that serve as labor market indicators and platform-specific signals in evaluating value in digital marketplaces. The study demonstrate labor signaling is changing in the era of digital work with implications for freelancers, platform developers and human resource management practices. It further challenges the traditional human capital development and raises questions about the role of traditional educational credentials in the future of digitally mediated work.
