Authors - Nitin Varshney, Harsh Nagar, Dhaval Pambhar, Krunal Vaghela Abstract - Federated Learning (FL) lets multiple devices work together to train a model while keeping data private. However, it has problems with managing trust, keeping the model’s integrity, and finding malicious participants in cloud environments. This article suggests TrustFLChain, a block-chain based FL framework that combines a decentralized trust management system with a scheme for collecting models based on deep learning. TrustFLChain uses a consortium blockchain to make sure that model updates can’t be changed, that participants’ trust is evaluated in a strong way using a hybrid identity-behavior trust model, and that the Proof-of-Trust (PoT) consensus mechanism works well. Trust scoring and the fraudulent detection updates are both automated by the use of smart contracts. Testing done on variety of datasets like the CIFAR-10, MNIST, and ImageNet shows that TrustFLChain is 6.5% more accurate, 38% faster to identify harmful updates, 28% faster at reaching a consensus, and 22% less energy-hungry than the current best methods. Significant improvements are supported by statistical studies (p < 0.01). TrustFLChain provides a flexible, safe, and private way to use trusted AI in cloud systems, perfect for a wide range of uses.
Authors - Harsh Nagar, Nitin Varshney, Twinkal Chavda, Krunal Vaghela Abstract - Machine learning (ML) produce a huge impact in the Fourth Industrial Revolution that will change everything. ML makes systems very precise and efficient in many sectors like as cyber security, healthcare, smart cities, and farming that is good for the environment. This article examines the different aspects and advancements in machine learning since 2021. It cover wide area which gives more details and fill gaps in crucial fields like ethical AI, real-time processing, energy efficiency, explainability, federated learning, and adversarial stability. We investigate supervised, unsupervised, semi-supervised, reinforcement, and deep learning algorithms, supported by solid theoretical frameworks, mathematical formulations, and empirical evaluations on extensive benchmark datasets (e.g., MNIST, CIFAR-10, NSL-KDD, UCI datasets). These contributions include deep study of performance evaluations, visual representations crafted with Python, and architectural frameworks for hybrid machine learning systems. These diverse results shows that how things work and implement in the real world and mark the difficulties like bias, scalability, and latency. The future will be about ML solutions that are precise, accurate, adaptable, and durable. This publication offers scholars and practitioners a technological framework that fosters innovation in the Fourth Industrial Revolution (4IR).
Authors - Mayur Rajendra Badiger, Tejasvi Bellubbi, Aditya Holla, Jayanth H T, Kavitha Devi CS Abstract - India, as the world’s largest democracy, faces the ongoing challenge of low voter turnout, with recent national elections recording a participation rate of just 65.79%. This issue is exacerbated by the existing electoral requirement that voters be physically present in their registered constituencies, effectively disenfranchising millions of migrant workers and citizens residing away from home. The proposed Smart and Secure Electronic Voting System addresses this critical gap by enabling inter-state voting, allowing eligible voters to cast ballots for their home constituencies from any location across the country. Built on an Arduino Uno platform, the system integrates biometric fingerprint authentication to ensure secure and unique voter identification, while an LCD interface displays constituency-specific candidate lists. Votes are securely transmitted and stored in a cloud-based database via a WiFi module, using the ThingSpeak server for real-time, tamper-resistant data management. The system prevents multiple voting attempts by locking out voters after a successful ballot submission, thereby upholding electoral integrity. By combining IoT technologies and robust biometric verification, this solution not only simplifies the voting process for the floating population but also strengthens the inclusivity, accessibility, and representativeness of the Indian democratic process.
Authors - Niranjan Kulkarni, Suma MN Abstract - This paper introduces an integrated model that couples financial literacy training with financial therapy to enhance financial knowledge as well as behavioural results. Basing its argument on existing literature, this study employs a mixed-methods methodology to assess the effectiveness of pilot programs that blend therapeutic interventions like cognitive-behavioural therapy (CBT) and financial counselling with standard financial education. The study unveils the need for an integrated financial education approach, providing a model for the development of more effective financial literacy programs that consider cognitive as well as affective dimensions of financial well-being.
Authors - W Saib, T Moodley Abstract - In modern football, smart recruitment is essential in building successful teams. Traditional scouting methods, though effective in identifying player mentalities, often fall short due to the team’s budget restrictions and the need to fit the head coach’s strategy. Teams frequently lose top targets to wealthier clubs, sign players who do not fit with the coach’s strategy and overlook affordable and stylistically similar players. To address this, we propose a novel player recommendation system, which first clusters players and then leverages the Graph Convolutional Networks (GCNs) ability to generate rich node embeddings from a player-similarity graph constructed via cosine distance. Using data from FBref, we refine the scope to only central midfielders and on the Graph Convolutional Network where the target variables are the labels from K-Means clustering of which we obtained 49% accuracy with all neighbours but a perfect 100% with shared-label aggregation introduced in this paper. It was observed that there were significant visual differences (obtained via t-SNE) in the clusters formed. Graph Convolutional Networks are primarily used for node classification on academic citation datasets, and the proposed method in this research outperformed on the Cora (88%) and Pubmed (84%) Datasets, respectively.
Authors - Precious Magodi, Tevin Moodley Abstract - Vision-based sign language recognition significantly advances communication within the deaf community, enhancing accessibility and inclusion for those who are deaf or hard of hearing. This paper presents a system developed for real-time recognition of South African Sign Language (SASL) using Google’s MediaPipe framework for spatial feature extraction and a long short-term memory (LSTM) network for temporal modelling.We utilise a subset of the ASL Citizen dataset, focusing on five classes: “SCHOOL,” “TIMEOUT,” “MORNING,” “THANK YOU,” and “I LOVE YOU,” which serve as proxies for SASL vocabulary. Keypoint sequences from both hands and body pose are extracted via MediaPipe and fed into a two-layer LSTM for classification. Trained with a Tensor- Flow TFRecord pipeline, our model achieves a test accuracy of 35.5% and highlights the challenges posed by limited data and variability among signers. This work demonstrates the potential of combining MediaPipe and LSTM for real-time sign recognition and emphasises the need for larger, language-specific datasets to improve accuracy.
Authors - Yuya Ieiri, Tengfei Shao, Osamu Yoshie Abstract - Social capital is recognized as an important concept for revitalizing commercial districts. However, the dynamics of social capital within these dis-tricts remain largely unexplored. As an initial step toward addressing this gap, this study conceptualizes changes in the social network of a commercial district community as shifts in social capital. Furthermore, it investigates the impact of information sharing within the community on this social capital. The study developed a framework to examine the relationship between social capital and so-cial networks. An empirical investigation involving ten participants and three experiments was conducted over three weeks. Results indicated that participants' subjective evaluations of store importance are influenced by the mode of information sharing within the community. Furthermore, the findings suggest that structured, repeated information sharing can strengthen interpersonal rela-tionships within the network. From a social-capital perspective, these findings imply that information sharing within interactive groups can potentially foster its development.
Authors - Nirel Marie M. Ibarra Abstract - The fast changes in front-end web development technologies have widened the gap between what schools teach and what businesses need in the Philippines. This paper offers the first systematic study of developer job postings in the country, using a new approach that combines web scraping, LLM-based classification, and statistical analysis. By examining hiring needs across small, mid-sized, and large companies, the study highlights apparent differences in technology preferences, range of skills, and business practices. The findings indicate that smaller firms look for adaptable generalists, while larger organizations focus on specialized skills and organized processes. There are also early signs of a demand for AI knowledge in various settings. These insights form a Minimum Viable Skill Set (MVSS) that includes technical skills like React, JavaScript/TypeScript, HTML, CSS, Git/GitHub workflows, and basic DevOps. It also includes the ability to adjust to different work environments. The results aim to guide curriculum changes in Philippine universities and act as a model for other ASEAN countries.
Authors - Suthawan Prukumpai, TzeHoung Lee, Patrick Ghali Abstract - Traditional correlation metrics like Pearson correlation capture only linear and symmetric relationships between variables, potentially obscuring meaningful dependencies in financial systems where nonlinearities and lead-lag effects are common. This paper introduces SVMBased Predictive Correlation (S2C), a novel approach that redefines correlation in terms of predictive power rather than mere association. S2C accommodates nonlinearities and directional structures within data, allowing practitioners to better gauge risk and properly size allocations. We develop the mathematical foundations of S2C within the context of kernel methods and reproducing kernel Hilbert spaces (RKHSs), then present an empirical study involving financial and macroeconomic time series with different update frequencies. Our analysis demonstrates that S2C can recover both linear and nonlinear relationships, is more robust to irregular sampling than standard correlation measures, and captures directionality by revealing how one variable can predict another. We conclude with implications for portfolio construction and risk management, particularly addressing the ”volatility washing” problem in portfolios containing illiquid alternative investments.
Authors - Patil Pratima, Deshpande Deepa Abstract - This paper presents a real-time thermal face detection and recogni-tion system based on an enhanced Multi-task Cascaded Convolutional Neural Network (MTCNN) framework. Unlike visible-light methods, thermal imaging introduces domain-specific challenges such as low spatial resolution, high noise, and intensity variance due to temperature fluctuations. To address these, we propose a dedicated preprocessing pipeline including normalization, contrast enhancement, and channel replication to adapt single-channel thermal images for CNN-based processing. The modified MTCNN is fine-tuned on thermal datasets to accurately detect facial regions and landmarks. Aligned faces are then processed through a thermal-optimized feature embedding network trained with tri-plet loss to produce identity-preserving descriptors. Recognition is performed us-ing a lightweight classifier over the feature space. The system is optimized for real-time performance using GPU acceleration and quantized inference. Experi-mental results on publicly available thermal face datasets demonstrate the effec-tiveness of our approach in terms of detection accuracy, recognition rate, and processing speed, making it suitable for surveillance and biometric applications under low-light or no-light conditions.
Authors - Asma Shahin, Deepa Deshpande Abstract - The prevalence of stress in modern life demands innovative solutions to monitor and mitigate its effects on individual well-being. This study explores stress detection through a robust dataset derived from wearable devices, analysing physiological signals, perceived stress scores, and task performance metrics. Given the widespread impact of stress across diverse populations, this research seeks to unravel the complexities of stress responses, paving the way for personalized interventions and enhancing our understanding of the physiological mechanisms underlying stress. On the Stress Predict dataset, the accuracy by Random Forest was the highest at 83% as compared to Logistic regression - 54% and SVM - 82%.
Authors - Ornurai Sangsawang Abstract - Industrial chemical emergencies, such as leaks, fires, and explosions, are major risks that affect life, property, the environment, and infrastructure security, especially in areas with high factory density. This research aims to develop a spatial emergency response station placement model by applying the Maximal Covering Location Problem (MCLP) model in two case studies: (1) the placement of the Industrial Emergency Response Center (I-ERC) using factory location data in Rayong Province, and (2) the placement of the Chemical Emergency Response Station (CERS) using historical chemical accidents. The results of the study compared the locations of 21 existing EMS stations and found that the MCLP model can significantly increase the coverage of the risk areas. In the I-ERC case, the coverage level increased from 87.09% to 99.46%, and in the CERS case, the risk points can be covered 100% with an appropriate number of stations. The results of this study can be applied to support strategic planning for emergency response systems in high-risk industrial areas.
Authors - Arnold Aguilar, Ronaldo Polancos Abstract - Multilingual IT contact centers serve as vital operational structures for global corporations, facilitating engagement with geographically dispersed clientele across diverse languages, temporal zones, and technical complexities. Excluding their strategic significance, the mechanisms for scheduling and task allocation within these contexts have remained inadequately investigated within the classical project scheduling literature. Conventional frameworks addressing the resource-constrained project scheduling problem (RCPSP) neglect to incorporate dynamic constraints such as agent fatigue, linguistic alignment, shift-based availability, or the demands of strategic escalation. This systematic literature review integrates more than twenty years of advancements in RCPSP, leveraging hybrid metaheuristic methodologies—specifically Genetic Algorithms (GA) and Constraint Programming (CP)—to propose an innovative trajectory tailored to multilingual service operations. A case study is employed to corroborate the proposed model, demonstrating its potential to enhance SLA compliance, agent workload distribution, and language-specific support efficacy. This research provides a dedicated conceptual and algorithmic foundation for multilingual support scheduling—an area that has been predominantly preoccupied from the mainstream RCPSP discourse—thereby establishing a basis for forthcoming empirical validation and implementation in high-demand support environments.
Authors - Kayode Daniel Akilo, Rhulani Maluleka, Abelwe Maluleka Abstract - The introduction of Artificial Intelligence (AI) in revolutionizing e-business environments has considerably automated operational efficiency with ease. The induction of AI has increased consumer engagement and business suc-cess across the e-commerce landscape. Concerning this, this research examines the application, operation, and various impacts and challenges of the adoption of artificial intelligence across e-commerce operations and platforms. Primarily focusing on Shopify as a case study, this allows for an effective and efficient analysis of AI adoption. This research utilizes SWOT analysis and comprehensive academic literature in evaluating AI environments and AI adoption, such as intelligent chatbots, personalization engines, and security systems. Through this analysis, this research extends to how these implementations transform both small and medium-sized enterprises (SMEs). The research identifies the barriers to implementation, strategic approaches in effective AI adoption, and factors that are to be attributed to success. Additionally, this research contributes to the competitive advantages through an established platform, such as Shopify, for a more grounded approach. This paper evaluates the AI adoption in the e-commerce landscape and addresses the economic infrastructure. Allowing effective solu-tions for SMEs in practical, technical, and ethical ways.
Authors - Balebetse H. Skele, Partson Paradza, Benita G. Zulch, Joseph A. Yacim Abstract - In South Africa, Section 2 of the Sectional Title Schemes Management Act of 2011 (the Act) requires developers to attend the First General Meeting (FGM) and present necessary documentation to the corporate body. However, non- compliance with standards and incomplete submissions by developers often lead to the postponement of the FGM. This results in financial burdens for the body corporate and the signing of long-term contracts without proper oversight. The impact of this non-compliance has not been investigated, highlighting the need for further investigation to improve handover procedures under the Act. Thus, this study aims to inform current policies and contribute to the literature on the handover process of residential sectional title properties. Adopting a case study approach, the research integrates mixed methods to collect data from seven sectional title schemes in the Western Cape and Gauteng Provinces. Findings reveal that developers' non-attendance at FGMs causes communication breakdowns with Body Corporate, rendering FGMs ineffective. This paper provides valuable insights for policy on sectional title purchases in South Africa. It also contributes to the litera-ture by examining the challenges in the handover processes of sectional title properties in South Africa.
