Learning predictive models from functional magnetic resonance imaging (fMRI) connectivity data is important for analyzing cognitive processes and neurological disorders. Most existing approaches adopt a single-task learning (STL) formulation, where models predict either cognitive task states or clinical diagnoses independently. This setting often limits generalization because related tasks share latent neural representations that remain unused. Multi-task learning (MTL) addresses this limitation by learning shared representations across tasks, enabling complementary information transfer and improved robustness. In this work, we propose the Brain Connectivity Transformer (BCT), a graph-based transformer architecture for joint cognitive task and disorder classification from fMRI connectivity graphs. BCT operates on region-of-interest (ROI) connectivity graphs and employs multi-head self-attention to model global interactions among brain regions. A dynamic ROI attention module identifies task- and disorder-relevant regions, while a weighted orthonormal clustering readout aggregates attention-refined features into cluster-level representations. Within the MTL framework, BCT jointly optimizes task and disorder objectives. Under subject independent evaluation, BCT achieves 85.13± 0.3% task accuracy and 95.30± 1.1% disorder accuracy, outperforming the strongest baseline by 4.51% and 7.08% , respectively. These results demonstrate the effectiveness of transformer-based MTL on connectivity graphs for joint cognitive task and disorder prediction. The source code is available at https://github.com/DDasChakladar/Brain-Connectivity-Transformer .

Research on continual learning (CL) mainly focuses on data represented in the Euclidean space, while research on graph-structured data is scarce. Furthermore, most graph learning models are tailored for static graphs. However, graphs usually evolve continually in the real world. Catastrophic forgetting also emerges in graph learning models when being trained incrementally. This leads to the need to develop robust, effective and efficient continual graph learning approaches. Continual graph learning (CGL) is an emerging area aiming to realize continual learning on graph-structured data. This survey is written to shed light on this emerging area. It introduces the basic concepts of CGL and highlights two unique challenges brought by graphs. Then it reviews and categorizes recent state-of-the-art approaches, analyzing their strategies to tackle the unique challenges in CGL. Besides, it discusses the main concerns in each family of CGL methods, offering potential solutions. Finally, it explores the open issues and potential applications of CGL.

This study presents a novel algorithm, FaRS, for single-source role similarity search, designed to capture nuanced topological features within graphs more effectively than existing methods. Traditional role-based similarity algorithms like RoleSim are proficient at identifying automorphic equivalences but often fail to distinguish nodes with structural differences despite their automorphic similarities. By incorporating a technique that utilizes the top Γ maximum similarity matching, FaRS enhances the fidelity of role similarity evaluations by considering a broader range of adjacency relationships. This approach not only ensures the accurate identification of automorphic and structural equivalences but also adheres to key mathematical properties such as uniqueness, symmetry, boundedness, and triangular inequality. We also introduce an accelerated variant of FaRS, named Opt_FaRS, which employs innovative computational strategies to improve efficiency, particularly in dynamic environments. Experimental validations on several real-world datasets demonstrate that FaRS and Opt_FaRS outperform standard benchmarks in both accuracy and computational speed, offering substantial improvements for applications in diverse domains like social network analysis and complex network management. This work contributes significant theoretical and practical advancements to the field of graph-based similarity search, laying a foundation for future explorations into dynamic graph analytics.

How to upgrade business processes to improve production efficiency is an ongoing concern in industrial research. While previous studies have extensively examined various prioritization schemes at each stage of the business process, there has been a lack of investigation into the financial resources required for their implementation. The attainment of sufficient and stable financial support necessitates stability in stock prices, making the control of significant volatility in stock markets a critical issue. This study examines the effectiveness of three design schemes of price limit policy, a prevalent policy that intends to control significant volatility in financial markets and stabilize the market. Utilizing a heterogeneous agent-based model that simulates trading agents' processes of updating strategies through genetic programming algorithms and incorporates specialized designs for price limit policies, this study demonstrates that an asymmetric limit policy-consisting solely of a lower price limit (without an upper price limit)-can significantly enhance market quality by achieving lower volatility, higher market liquidity and better price effectiveness. Furthermore, we investigate the applicable conditions of asymmetric price limits. The findings suggest that an extremely restrictive limit range could lead to volatility spillover, while a 10% range is deemed appropriate for achieving better efficiency. Additionally, the asymmetric price limit mechanism has the potential to significantly reduce market volatility by up to 12.5% in volatile, low liquidity, and low price efficiency markets, which aligns with the declining range from bubble-crash periods to stable periods in the Chinese stock market. These results are further supported by sensitivity analysis.

Enterprise security operations centers evaluate millions of events daily, yet $85-95 \%$ are false positives or innocuous activities, causing analyst overload and delayed threat identification and response. Endpoint Detection identifies device vulnerabilities, while User Behavior Analytics detects anomalies in user behavior. They operate independently with limited detection capabilities. The Distributed Edge Fusion Framework (DEFF) enables network-edge agents to correlate various signals without centralizing raw data. Edge processors standardize vulnerability counts, patch age, and operating system risk and behavioral indicators (logon velocity, geographic disparities, and temporal entropy) before sending aggregated scores to a central system using weighted sigmoid activation functions. Central ensemble classifiers, utilizing Random Forest, XGBoost, and logistic regression, outperform community benchmarks $(0.84-0.87)$ by 6 % to 10 %. DEFF reduces false positives by 72 %, data transmission volume by 87 %, and detection delay by 92 %. GDPR and CCPA-compliant architecture protects privacy and enables precise detection. 13.6 million event reports demonstrate business system capability using real class distributions. DEFF shows that decentralized edge-level integration of diverse security signals meets operational, latency, and legal constraints better than centralized ensemble learning.

This paper provides a cloud-based academic recommender system that integrates multiple heterogeneous big data academic resources to deliver personalized recommendations within the academic networks. The proposed system introduces a hybrid recommendation mechanism combining content-based and collaborative filtering within a graph-based relationship model.

The increasing population and emerging business opportunities have led to a rise in consumer spending. Consequently, global credit card companies, including banks and financial institutions, face the challenge of managing the associated credit risks. It is crucial for these institutions to accurately classify credit card customers as “good” or “bad” to minimize capital loss. This research investigates the approaches for predicting the default status of credit card customer via the application of various machine-learning models, including neural networks, logistic regression, AdaBoost, XGBoost, and LightGBM. Performance metrics such as accuracy, precision, recall, F1 score, ROC, and MCC for all these models are employed to compare the efficiency of the algorithms. The results indicate that XGBoost outperforms other models, achieving an accuracy of 99.4%. The outcomes from this study suggest that effective credit risk analysis would aid in informed lending decisions, and the application of machine-learning and deep-learning algorithms has significantly improved predictive accuracy in this domain.

Federated learning enables model training for the consumer-driven Internet of Things (IoT) in a distributed manner without violating individual privacy. Several secure aggregation protocols have been proposed for large-scale federated learning models in IoT scenarios. However, the communication and computational overheads grow quadratically with the number of clients, which becomes a significant obstacle to these secure aggregation protocols. To address this problem, some work utilizes deterministic graphs of logarithmic degrees, such as the Harary graph or Erd & odblac;s-Renyi graph, instead of the complete communication graph. The graph generated under the given fixed conditions is unique and invariant throughout the federated learning process. In this paper, we propose SparsiFL, a graph sparsification-based secure aggregation protocol for federated learning, which significantly reduces communication and computational overhead while maintaining correctness and privacy. SparsiFL takes a complete graph as input and formulates the optimization problem as an uncertain graph sparsification task, which reduces the number of edges and redistributes the probabilities attached to them. In the process, SparsiFL also preserves the underlying structure. The graph can accurately and efficiently approximate the secure-sharing task in secure aggregation. Theoretical analysis shows correctness and privacy. Experiments show that SparsiFL reduces the communication and computational overheads up to 6.10x and 3.16x as compared to other related approaches.

Federated learning (FL) is revolutionizing mobile computing and IoT development by enhancing data privacy. However, restricted computational and communication resources and the statistical variability of data stored on devices present substantial obstacles to ongoing progress in FL. We introduce EarlyBirdFL, a novel FL framework that leverages an Early-Bird Ticket-inspired pruning and masking technique for efficient training and communication in federated settings. EarlyBirdFL enables each client to achieve fast local training by identifying efficient subnetworks early in the training process, communicating only these pruned networks between the server and the client. Unlike classical personalized FL, in which the client-side model learns differences, EarlyBirdFL allows each client to identify these efficient subnetworks using a mask metric quickly. Experimental results demonstrate that EarlyBirdFL outperforms traditional computation time and accuracy methods, achieving a 1.53-4.98 times speedup and 1.01-1.15 times higher accuracy. Furthermore, EarlyBirdFL remains stable even when its parameters are adjusted and performs well in different non-IID environments, maintaining or surpassing the performance of other methods. This approach combines elements of early efficient subnetwork identification, pruning, masking, and personalized federated learning to address the unique challenges of FL.

Smart technologies have demonstrated striking outcomes regarding the early diagnosis of diseases and the delivery of the necessary healthcare in the last decade. However, by emphasizing the core fundamentals of social justice and sustainability, together with digitalization and smart technologies that predicate raising productivity and flexibility, Industry 5.0 has proven to achieve more efficient results. Industry 5.0 technologies provide more intelligent ways for human employees and higher efficiency development while also improving safety and performance in many applications. In this research, the contribution is focused on the healthcare and how Industry 5.0 technologies demonstrate several advantages for the healthcare sector, starting with automated and precise disease prediction, moving on to aiding medical personnel in continual surveillance and monitoring and concluding with successful digital automation of smart equipment. The objective of this study is to apply a hybrid multi-criteria decision-making approach under a neutrosophic environment to evaluate the advantages of industry 5.0 technologies in the healthcare sector. Industry 5.0 primary value is to reach human-centric, sustainable, and resilient industries. While Industry 5.0 technologies sub-values regarding the healthcare sector are determined and distinguished according to the 3-main values mentioned previously based on literature. The methodologies applied in this study are: The Analytical Hierarchy approach (AHP) evaluates the main values and sub-values. Subsequently, the effectiveness of industry 5.0 technologies according to their values to the healthcare sector are ranked by Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS). The approach is constructed under uncertainty based on a neutrosophic environment to achieve accuracy in the evaluation process. The results show that the most influential technology in healthcare are AI and cloud computing, while nano-technology, drone technology, and robots are at the end of the ranking. While validating the suggested technique, outcome comparisons were carried out to demonstrate the benefits of the methodologies. A sensitivity study indicates that adjusting the weightings of the sub-values has no significant effect on the ranking of technologies.

Customer churn is a significant concern, and the telecommunications industry has the largest annual churn rate of any major industry at over 30%. This study examines the use of ensemble learning models to analyze and forecast customer churn in the telecommunications business. Accurate churn forecasting is essential for successful client retention initiatives to combat regular customer churn. We used innovative and improved machine learning methods, including Decision Trees, Boosted Trees, and Random Forests, to enhance model interpretability and prediction accuracy. The models were trained and evaluated systematically by using a large dataset. The Random Forest model performed best, with 91.66% predictive accuracy, 82.2% precision, and 81.8% recall. Our results highlight how well the model can identify possible churners with the help of explainable AI (XAI) techniques, allowing for focused and timely intervention strategies. To improve the transparency of the decisions made by the classifier, this study also employs explainable artificial intelligence methods such as LIME and SHAP to illustrate the results of the customer churn prediction model. Our results demonstrate how crucial it is for customer relationship managers to implement strong analytical tools to reduce attrition and promote long-term economic viability in fiercely competitive marketplaces. This study indicates that ensemble learning models have strategic implications for improving consumer loyalty and organizational profitability in addition to confirming their performance.

China's reform and opening-up policies have prioritized technological advancement, with technological SMEs driving employment and economic growth. Despite their significance, these SMEs face substantial financing and operational risks due to inadequate credit measurement tools. This study reviews the historical financing challenges of technological SMEs since the 1980s, summarizes their current credit risk status, and compares four modern credit risk models: Credit Metrics, Credit Risk+, Credit Portfolio View, and KMV. We propose a pioneering KMV Strategy for real-time risk analysis, contributing to accurate credit metrics for these SMEs. Finally, we suggest policies for managing their credit risks through prevention, control, and governance.

Outdoor images taken in haze usually exhibit contrast reduction, color distortion, and detail loss. Removing the haze from a given image is a tough issue owing to its highly ill-posed property. To restore the haze-free image effectively, we develop an unsupervised dehazing method using patch-line and fuzzy clustering-line priors in this article. The method obtains the dehazed image by inversely solving the atmospheric scattering model, which involves in estimating two key parameters, including atmospheric light and scene transmission. First, the orientation of atmospheric light is achieved by using a patch-line prior. Then, a quadtree subspace hierarchical searching scheme is designed to get the magnitude by calculating the differences between the mean and variance of each component for local regions. Besides, a fuzzy clustering-line prior combined with a guided filtering is proposed to estimate the scene transmission for each pixel. The proposed method can obtain the dehazed image directly without any training process and achieve much better performance than many existing ones with less space and time cost.

Sentiment analysis has demonstrated its value in a range of high-stakes domains. From financial markets to supply chain management, logistics, and technology legitimacy assessment, sentiment analysis offers insights into public sentiment, actionable data, and improved decision forecasting. This study contributes to this growing body of research by offering a novel multi-view deep learning approach to sentiment analysis that incorporates non-textual features like emojis. The proposed approach considers both textual and emoji views as distinct views of emotional information for the sentiment classification model, and the results acknowledge their individual and combined contributions to sentiment analysis. Comparative analysis with baseline classifiers reveals that incorporating emoji features significantly enriches sentiment analysis, enhancing the accuracy, F1-score, and execution time of the proposed model. Additionally, this study employs LIME for explainable sentiment analysis to provide insights into the model's decision-making process, enabling high-stakes businesses to understand the factors driving customer sentiment. The present study contributes to the literature on multi-view text classification in the context of social media and provides an innovative analytics method for businesses to extract valuable emotional information from electronic word of mouth (eWOM), which can help them stay ahead of the competition in a rapidly evolving digital landscape. In addition, the findings of this paper have important implications for policy development in digital communication and social media monitoring. Recognizing the importance of emojis in sentiment expression can inform policies by helping them better understand public sentiment and tailor policy solutions that better address the concerns of the public.

Cyber threat intelligence (CTI) is widely recognized as an important area in cybersecurity but it remains an area showing silos and reserved for large organizations. For an area whose strength is in open and responsive sharing, we see that the generation of feeds has a small scale, is secretive, and is nearly always from specialized businesses that have a commercial interest in not publicly sharing insights at a speed where it could be effective in raising preparedness or stopping an attack. This article has three purposes. First, we extensively review the state and challenges of open, crowd-sourced CTI, with a focus on the perceived barriers. Second, having identified that confidentiality (in multiple forms) is a key barrier, we perform a confidentiality threat analysis of existing sharing architectures and standards, including reviewing circa one million of real-world feeds between 2014 and 2022 from the popular open platform MISP toward quantifying the inherent risks. Our goal is to build the case that, either by redesigning sharing architectures or simply performing simple sanitization of shared information, the confidentiality argument is not as strong as one may have presumed. Third, after identifying key requirements for open crowd-based sharing of CTI, we propose a reference (meta-) architecture. Managerial Relevance—CTI is widely recognized as a key advantage toward cyber resilience in its multiple dimensions, from business continuity to reputation/regulatory protection. Furthermore, as we review in this article, there are strong indications that the next generation of approaches to cybersecurity will be centered on CTI. Whereas CTI is an established business area, we see little adoption, closed communities, or high costs that small businesses cannot afford. For an area that, intuitively, should be open, as velocity and accuracy of information is crucial, we shed light on why we have no significant open, crowd-sourced CTI. In other words, why is usage so lacking? We identify reasons and deconstruct unclear and unhelpful rationales by looking at a wide range of literature (research and professional) and an analysis of nearly ten years of open CTI data. Our findings from current data indicate two types of reasons. One, and dominant, is unhelpful perceptions (e.g., confidentiality), and another stems from market factors (e.g., “free-riding”) that need collective movement as no single player may be able to break the cycle. After looking at motivations and barriers, we review existing technologies, elicit requirements, and propose a high-level open CTI sharing architecture that could be used as a reference for practitioners.

This paper contributes to rapid prototyping for Aboleth, a mobile application designed as a virtual character sheet for Dungeons and Dragons 5th edition. We propose augmenting an API framework, including details of the general design process and crucial details regarding the specific implementation of the project. The prototype is developed using Computer-Assisted Designs and Xamarin. Software engineering is integrated into game design to improve Human-Computer Interaction and eventually enhance user experiences. Additionally, we discuss the relevance of using such a tool to achieve rapid prototyping for enterprise mobile applications and Metaverse and explore opportunities to improve the project design.

The control of credit risk is an important topic in the development of supply chain finance. Financial service providers should distinguish between low- and high-quality customers to predict credit risk accurately. Proper management of credit risk exposure contributes to the long-term viability and profitability of banks, systemic stability, and efficient capital allocation in the economy. Moreover, it benefits the development of supply chain finance. Supply chain finance offers convenient loan transactions that benefit all participants, including the buyer, supplier, and bank. However, poor credit risk management in supply chain finance may cause losses for finance providers and hamper the development of supply chain finance. Machine learning algorithms have significantly improved the accuracy of credit risk prediction systems in supply chain finance. However, their lack of interpretability or transparency makes decision-makers skeptical. Therefore, this study aims to improve AI transparency by ranking the importance of features influencing the decisions made by the system. This study identifies two effective algorithms, Random Forest and Gradient Boosting models, for credit risk detection. The factors that influenced the decision of the models to make them transparent are explicitly illustrated. This study also contributes to the literature on explainable credit risk detection for supply chain finance and provides practical implications for financial institutions to inform decision making.