Knowledge and Information Systems最新文献

Credit risk, stemming from the failure of a contractual party, is a significant variable in financial institutions. Assessing credit risk involves evaluating the creditworthiness of individuals, businesses, or entities to predict the likelihood of defaulting on financial obligations. While financial institutions categorize consumers based on creditworthiness, there is no universally defined set of attributes or indices. This research proposes Range control-based class imbalance and Optimized Granular Elastic Net regression (ROGENet) for feature selection in credit risk assessment. The dataset exhibits severe class imbalance, addressed using Range-Controlled Synthetic Minority Oversampling TEchnique (RCSMOTE). The balanced data undergo Granular Elastic Net regression with hybrid Gazelle sand cat Swarm Optimization (GENGSO) for feature selection. Elastic net, ensuring sparsity and grouping for correlated features, proves beneficial for assessing credit risk. ROGENet provides a detailed perspective on credit risk evaluation, surpassing conventional methods. The oversampling feature selection enhances the accuracy of minority class by 99.4, 99, 98.6 and 97.3%, respectively.

This work presents an approach for distributed and contextualized reasoning in multi-agent systems, considering environments in which agents may have incomplete, uncertain and inconsistent knowledge. Knowledge is represented by defeasible logic with mapping rules, which model the capability of agents to acquire knowledge from other agents during reasoning. Based on such knowledge representation, an argumentation-based reasoning model that enables distributed building of reusable argument structures to support conclusions is proposed. Conflicts between arguments are resolved by an argument strength calculation that considers the trust among agents and the degree of similarity between knowledge of different agents, based on the intuition that greater similarity between knowledge defined by different agents implies in less uncertainty about the validity of the built argument. Contextualized reasoning is supported through sharing of relevant knowledge by an agent when issuing queries to other agents, which enable the cooperating agents to be aware of knowledge not known a priori but that is important to reach a reasonable conclusion given the context of the agent that issued the query. A distributed algorithm is presented and analytically and experimentally evaluated asserting its computational feasibility. Finally, our approach is compared to related work, highlighting the contributions presented, demonstrating its applicability in a broader range of scenarios, and presenting perspectives for future work.

In recent years, neural network search has been utilized in designing effective heterogeneous graph neural networks (HGNN) and has achieved remarkable performance beyond manually designed networks. Generally, there are two mainstream design manners in heterogeneous graph neural architecture search (HGNAS). The one is to automatically design a meta-graph to guide the direction of message-passing in a heterogeneous graph, thereby obtaining semantic information. The other learns to design the convolutional operator aiming to enhance message extraction capabilities to handle the diverse information in a heterogeneous graph. Through experiments, we observe a strong interdependence between message-passing direction and message extraction, which has a significant impact on the performance of HGNNs. However, previous HGNAS methods focus on one-sided design and lacked the ability to capture this interdependence. To address the issue, we propose a novel perspective called heterogeneous message-passing mechanism for HGNAS, which enables HGNAS to effectively capture the interdependence between message-passing direction and message extraction for designing HGNNs with better performance automatically. We call our method heterogeneous message-passing mechanisms search (HMMS). Extensive experiments on two popular tasks show that our method designs powerful HGNNs that have achieved SOTA results in different benchmark datasets. Codes are available at https://github.com/HetGNAS/HMMS.

Semi-supervised learning is a promising approach to dealing with the problem of insufficient labeled data. Recent methods grouped into paradigms of consistency regularization and pseudo-labeling have outstanding performances on image data, but achieve limited improvements when employed for processing textual information, due to the neglect of the discrete nature of textual information and the lack of high-quality text augmentation transformation means. In this paper, we propose the novel SeqMatch method. It can automatically perceive abnormal model states caused by anomalous data obtained by text augmentations and reduce their interferences and instead leverages normal ones to improve the effectiveness of consistency regularization. And it generates hard artificial pseudo-labels to enable the model to be efficiently updated and optimized toward low entropy. We also design several much stronger well-organized text augmentation transformation pipelines to increase the divergence between two views of unlabeled discrete textual sequences, thus enabling the model to learn more knowledge from the alignment. Extensive comparative experimental results show that our SeqMatch outperforms previous methods on three widely used benchmarks significantly. In particular, SeqMatch can achieve a maximum performance improvement of 16.4% compared to purely supervised training when provided with a minimal number of labeled examples.

Attributed graph clustering or community detection which learns to cluster the nodes of a graph is a challenging task in graph analysis. Recently contrastive learning has shown significant results in various unsupervised graph learning tasks. In spite of the success of graph contrastive learning methods in self-supervised graph learning, using them for graph clustering is not well explored. In this paper, we introduce a contrastive learning framework for learning clustering-friendly node embedding. We propose Gaussian mixture information maximization which utilizes a mutual information maximization approach for node embedding. Meanwhile, in order to have a clustering-friendly embedding space, it imposes a mixture of Gaussians distribution on this space. The parameters of the contrastive node embedding model and the mixture distribution are optimized jointly in a unified framework. Experiments show that our clustering-directed embedding space can enhance clustering performance in comparison with the case where community structure of the graph is ignored during node representation learning. The results on real-world datasets demonstrate the effectiveness of our method in community detection.

Multi-Attribute Similarity Join represents an important task for a variety of applications. Due to a large amount of data, several techniques and approaches were proposed to avoid superfluous comparisons between entities. One of these techniques is denominated Index Tree. In this work, we proposed an adaptive version (Adaptive Index Tree) of the state-of-the-art Index Tree for multi-attribute data. Our method selects the best filter configuration to construct the Adaptive Index Tree. We also proposed a reduced version of the Index Trees, aiming to improve the trade-off between efficacy and efficiency for the Similarity Join task. Finally, we proposed Filter and Feature selectors designed for the Similarity Join task. To evaluate the impact of the proposed approaches, we employed five real-world datasets to perform the experimental analysis. Based on the experiments, we conclude that our reduced approaches have produced superior results when compared to the state-of-the-art approach, specially when dealing with datasets that present a significant number of attributes and/or and expressive attribute sizes.

Class imbalance has been widely accepted as a significant factor that negatively impacts a machine learning classifier’s performance. One of the techniques to avoid this problem is to balance the data distribution by using sampling-based approaches, in which synthetic data is generated using the probability distribution of the classes. However, this process is sensitive to the presence of noise in the data, and the boundaries between the majority class and the minority class are blurred. Such phenomena shift the algorithm’s decision boundary away from the ideal outcome. In this work, we propose a hybrid framework for two primary objectives. The first objective is to address class distribution imbalance by synthetically increasing the data of a minority class, and the second objective is, to devise an efficient noise reduction technique that improves the class balance algorithm. The proposed framework focuses on removing noisy elements from the majority class, and by doing so, provides more accurate information to the subsequent synthetic data generator algorithm. To evaluate the effectiveness of our framework, we employ the geometric mean (G-mean) as the evaluation metric. The experimental results show that our framework is capable of improving the prediction G-mean for eight classifiers across eleven datasets. The range of improvements varies from 7.78% on the Loan dataset to 67.45% on the Abalone19_vs_10-11-12-13 dataset.

Knowledge Discovery and Management (KDM) encompasses a comprehensive process and approach involving the creation, discovery, capture, organization, refinement, presentation, and provision of data, information, and knowledge with a specific goal in mind. At the core, Knowledge Management and Artificial Intelligence (AI) revolve around knowledge itself. AI serves as the mechanism enabling machines to obtain, acquire, process, and utilize information, thereby executing tasks and uncovering knowledge that can be shared with people to enhance strategic decision-making. While conventional methods play a role in the KDM process, incorporating intelligent approaches can further enhance efficiency in terms of time and accuracy. Intelligent techniques, particularly soft computing approaches, possess the ability to learn in any environment by leveraging logic, reasoning, and other computational capabilities. These techniques can be broadly categorized into Learning algorithms (Supervised, Unsupervised, and Reinforcement), Logic and Rule-Based algorithms (Fuzzy Logic, Bayesian Network, and CBR-RBR), Nature-inspired algorithms (Genetic algorithm, Particle Swarm Optimization, and Ant Colony Optimization), and hybrid approaches that combine these algorithms. The primary objective of these intelligent techniques is to address the day-to-day challenges faced by rural and smart digital societies. In this study, the authors extensively investigated various intelligent computing methods (ICMs) specifically relevant to distinct problems, providing accurate and reasonable knowledge-based solutions. The application of both single ICMs and combined ICMs was explored to solve domain-specific problems, and their effectiveness was analyzed and discussed. The results indicated that combined ICMs exhibited superior efficiency compared to single ICMs. Furthermore, the authors conducted an analysis and comparison of ICMs based on their application domain, parameters, methods/algorithms, efficiency, and acceptable outcomes. Additionally, the authors identified several problem scenarios that can be effectively resolved using intelligent techniques.

This paper introduces a new dissimilarity measure between two discrete and finite probability distributions. The followed approach is grounded jointly on mixtures of probability distributions and an optimization procedure. We discuss the clear interpretation of the constitutive elements of the measure under an information-theoretical perspective by also highlighting its connections with the Rényi divergence of infinite order. Moreover, we show how the measure describes the inefficiency in assuming that a given probability distribution coincides with a benchmark one by giving formal writing of the random interference between the considered probability distributions. We explore the properties of the considered tool, which are in line with those defining the concept of quasimetric—i.e. a divergence for which the triangular inequality is satisfied. As a possible usage of the introduced device, an application to rare events is illustrated. This application shows that our measure may be suitable in cases where the accuracy of the small probabilities is a relevant matter.

Every year the volume of information is growing at a high rate; therefore, more modern approaches are required to deal with such issues efficiently. Distributed systems, such as Apache Spark, offer such a modern approach, resulting in more and more machine learning models, being adapted into using distributed logic. In this paper, we propose a classification model, based on Bayesian Networks (BNs), that utilizes the distributed environment of Apache Spark using the Dataframes paradigm. This model can exploit any user-provided directed acyclic graph (DAG) that portrays the dependencies between the features of a dataset to estimate the parameters of the conditional probability distributions associated with each node in the graph to make accurate predictions. Moreover, in contrast with the majority of implementations that are only able to handle discrete features, it is also capable of efficiently handling continuous features by calculating the Gaussian probability density function.