In recent years, the tabular generative model has emerged as a major research area across finance applications (e.g., fraud detection), especially under circumstances where original data are unavailable due to financial regulations such as the General Data Protection Regulation (GDPR). However, there is a notable scarcity of research on tabular generative models that directly handle the three characteristics of Financial Fraud Detection (FFD) data toward more effective FFD: (a) extreme class imbalance, (b) high data sparsity, and (c) a large number of attributes with non-normal distribution. Through a preliminary study, we confirmed that the performance of existing tabular generative models declines as the intensity of the three characteristics increases. To address these challenges, we propose novel graph-theoretical tabular generative models for tabular data generation, which are specialized for FFD. More specifically, we implement three variants of our method— SeparateGGM , SignedGGM , and HybridGGM —which are tailored to capture class-wise structure, signed inter-class relationships, and their integration, respectively. Our models jointly leverage several graph-theoretical methodologies, including graph neural network-based feature augmentation, graph topology and connectivity analysis, and new graph centrality indicators that we design in this work. Especially, through the graph analysis, we empirically find that certain hyperparameter values (i.e., positive and negative K ratios in separate and signed K-NN graphs), which make the related graph measurements approximately the highest points, consistently yield the highest detection accuracy, thus facilitating the selection of effective graph settings. Experimental results show that our method outperforms seven different competing methods across three popular FFD benchmarks and one private dataset. Finally, we plan to deploy SignedGGM , the best efficient variant of our method, and HybridGGM-A , the best-performing variant of our method, in a real-world financial application in the Financial Security Institute (FSI), one of South Korea’s financial government agencies.

The development of effective fraud detection systems (FDS) is hindered by strict data privacy regulations that prevent centralized data sharing. Federated learning (FL) has emerged as a promising alternative, enabling collaborative model training without exposing sensitive data. While FL has been explored in the healthcare domain, research on its application to financial fraud detection remains relatively limited. Specifically, FL research on real-world banking fraud types-with detailed customer, account, and transaction data-remains underexplored. We present the first empirical study of federated gradient boosting models for financial fraud detection in the banking sector, motivated by their superior performance over deep learning models on tabular fraud data. We evaluate and compare four representative federated gradient boosting models using both a private multi-fraud banking dataset from the Financial Security Institute (FSI) and a publicly available banking dataset, under various scenarios. Key findings include the consistent superiority of FedXGBBagging (a federated gradient boosting model), general vulnerability to data quantity skew, performance instability under bank join/dropout, and limitations in detecting localized banking fraud types such as ATM skimming. The findings from our empirical study highlight challenges and design considerations for deploying FL-based FDSs in the banking sector.

Notably, current intelligent transportation systems rely heavily on accurate traffic forecasting and swift inference provision to make timely decisions. While Graph Convolutional Networks (GCNs) have shown benefits in modeling complex traffic dependencies, the existing GCN-based approaches cannot fully extract and fuse multi-granular spatiotemporal features across various spatial and temporal scales sufficiently in a complete manner, proven to yield less accurate results. As extracting multi-granular features across scales has been a promising strategy across domains such as computer vision, natural language processing, and time-series forecasting, pioneering studies have attempted to leverage a similar mechanism for spatiotemporal traffic data mining. However, additional feature extraction branches introduced in prior studies critically increased model complexity and extended inference time, making it challenging to provide fast forecasts. In this paper, we propose MultiGran-STGCNFog, an efficient fog distributed inference system with a novel traffic forecasting model that employs multi-granular spatiotemporal feature fusion on generated dynamic traffic graphs to fully capture interdependent traffic dynamics. The proposed scheduling algorithm GA-DPHDS, optimizing layer execution order and layer-device scheduling scheme simultaneously, contributes to considerable inference throughput improvement by coordinating heterogeneous fog devices in a pipelined manner. Extensive experiments on real-world datasets demonstrate the superiority of the proposed method over selected GCN baselines.

There are an increasing number of public Internet of Things (IoT) devices installed in urban environments, with which users can perform a wide variety of tasks. Owing to the nature of public spaces, such IoT devices must support groups of users rather than just individuals. However, because the type and quality of IoT devices in public environments varies, it may be difficult for groups of users to recognize the opportunities to perform tasks. Moreover, group users are often new to a certain public place, and have not previously performed tasks in IoT-enriched public spaces. In this paper, we propose a two-phase task recommendation approach for groups of IoT users in public environments. In the first phase, we employ a random walk with restart (RWR) algorithm to overcome the problem of sparse historical data for the performance of user tasks in public IoT environments. The second phase predicts a set of operations (IoT device functionalities) that are most appropriate for each candidate task. In this phase, to more effectively predict IoT operations for a user task we consider the contextual semantics of users via a classification model. We evaluate our approach using real-world datasets collected from practical IoT testbed environments. In addition, we show that an appropriate set of task operations can be predicted effectively by considering task types and contextual semantics.

When a user of a computer system is performing more than one task at the same time, her error rate increases drastically. In any system this is a critical issue, since the goals of the tasks are not likely to be met. In that sense, the purpose of mental workload assessment is to estimate the mental demand of tasks to take action according to that, avoiding execution errors. In this paper we study two techniques of mental workload assessment, physiological signals and simulation models of mental behavior with the ACT-R cognitive architecture. The contributions of this study are in two folds: validate a positive correlation among physiological and simulated data and, to develop a supervised model of classification with a cost-sensitive feature selection algorithm using the ACT-R simulated data as an input of the model. Results show a positive, significant correlation among the two data sources, and that the model selects features of less cost and classify better than a baseline approach with 93.1% accuracy in average.

The emergence of computing power and the abundance of data have made it possible to assist human decisions, especially in the stock markets, in which the ability to predict future values would lower the risk of investing. In this paper, we present a new approach for identifying the predictive power of public emotions extracted from various sections of daily news articles on the movements of stock market indices. The approach utilizes the results of a lexicon emotion analysis conducted on crowd-annotated news to extract various types of public emotions from daily news articles. We also propose a model and an analysis method to score news articles regarding public emotions, and to identify which news sections and emotions cause movements in a stock market index. The results of an experiment conducted with 24,763 news articles show that some types of public emotions are significantly correlated with changes in the trading volume and the closing price of a stock market.

The evolution of the Web has allowed the generation of several platforms for collaborative work. One of the main contributors to these advances is the Open Source initiative, in which projects are boosted to a new level of interaction and cooperation that improves their software quality and reliability. In order to understand how the group of contributors interacts with the software under development, we propose a novel methodology that adapts Lotka-Volterra-based biological models used for host-parasite interaction. In that sense, we used the concept mutualism from social parasites. Preliminary results based on experiments on the Github collaborative platform showed that Open Source phenomena can be modeled as a mutualistic system, in terms of the evolution of the population of developers and repositories.

The proactive and spontaneous delivery of Web services for users on the move can lead to the depletion of their cognitive resources, affecting the normal processes of their physical activities. This is due to the competition for limited cognitive resources between the human-computer interactions required by Web services and the users' physical activities. This paper introduces a mechanism for binding and scheduling Web services based on an assessment of this competition for users on the move. The proposed approach is built on two theories from cognitive psychology. This mechanism is realized by a descriptive model of activities and Web services which is enriched with a cognitive layer. A computational model uses this description to assess the degree of the demand for cognitive resources by both the physical activities and the Web services. Additionally, a Web services coordination mechanism based on this level of demand, the principle of progressive disclosure, and the temporal concurrency of Web services ensures less cognitively taxing Web service compositions.

The vision of spontaneously delivering ubiquitous services to mobile users stands on the tradition of reflecting functional and non-functional requirements. Nevertheless, existing approaches do not reason in advance about the potential conflicts of mobility activities with human-computer interaction based tasks. These conflicts may burden the user with unexpected interruptions that overload the central human-cognitive capacity. This paper introduces a novel cognitive engineering mechanism to optimize service functionality co ordinations during runtime in accordance with situational demands of cognitive resources. We base our resource aware approach for service coordination optimization on two theories from cognitive psychology - the human-processing system theory of Navon and the multiple resource theory of Wickens. On top of this psychological background, we introduce a specific mechanism for varying service co ordinations.

Dynamic Web content provides us with time-sensitive and continuously changing data. To glean up-to-date information, users need to regularly browse, collect and analyze this Web content. Without proper tool support this information management task is tedious, time-consuming and error prone, especially when the quantity of the dynamic Web content is large, when many information management services are needed to analyze it, and when underlying services/network are not completely reliable. This paper describes a multi-level, lifecycle (design-time and run-time) coordination mechanism that enables rapid, efficient development and execution of information management applications that are especially useful for processing dynamic Web content. Such a coordination mechanism brings dynamism to coordinating independent, distributed information management services. Dynamic parallelism spawns/merges multiple execution service branches based on available data, and dynamic run-time reconfiguration coordinates service execution to overcome faulty services and bottlenecks. These features enable information management applications to be more efficient in handling content and format changes in Web resources, and enable the applications to be evolved and adapted to process dynamic Web content.