ABSTRACT Class imbalance and class overlapping are two prevalent data traits that collectively impair the generalisation performance of credit classification models. For this superposition trait, conventional sampling techniques or remedial measures either amplify marginal noise or irreversibly discard useful majority information. To address these issues, a hybrid sampling method integrating a decision tree (DT) and a subregional synthetic minority oversampling technique (SR‐SMOTE) algorithm is proposed. On the one hand, a decision tree‐based undersampling (DTU) algorithm is proposed, which innovatively embeds classifier training into the sampling process to iteratively identify and eliminate hard‐to‐classify samples from the majority class. This can effectively retain the key information of the majority class and alleviate the problem of data imbalance. On the other hand, an adaptive SR‐SMOTE algorithm is developed, which partitions the minority class boundary into subregions based on local class overlap density and performs weighted oversampling within each subregion. This focuses on the most ambiguous areas and mitigates the risk of generating overlapping synthetic samples, thereby reducing inter‐class overlap in imbalanced data. The proposed method is validated on four imbalanced credit datasets. Experimental results show that it outperforms benchmark models. Robustness tests with GBDT, KNN, and LogR confirm consistent performance gains, with GBDT achieving the best results. This suggests the proposed approach is effective for credit classification with imbalanced and overlapping data.

This paper proposes a two-phase unilateral rule extraction method based on a multiobjective genetic algorithm (URE-MOGA) to solve the challenges of sparse categorical features and model interpretability in credit risk classification tasks. In this method, rule encoding is employed to represent the sparse categorical features without increasing data dimensionality. The Pareto dominance theory is utilized to balance the trade-off between model accuracy and interpretability. Additionally, a diversity function and unilateral rule learning strategy are designed to facilitate the acquisition of reasonable rules and classifiers by the URE-MOGA model. Furthermore, three publicly available datasets, along with six derived datasets, are used to validate the feasibility of the proposed URE-MOGA model. Experimental results demonstrate that the proposed model outperforms baseline methods in terms of accuracy, interpretability, and robustness. The adoption of a multiobjective approach enables the URE-MOGA model to extract more concise and comprehensible classification rules while ensuring high accuracy levels. Moreover, both the unilateral rule learning strategy and diversity function play vital roles in enhancing classifier accuracy and robustness within the URE-MOGA model. Overall, the proposed URE-MOGA model provides a novel insight into categorical feature sparsity handling interpretable modeling for credit risk classification tasks.

In credit fraud detection practice, certain fraudulent transactions often evade detection because of the hidden nature of fraudulent behavior. To address this issue, an increasing number of positive-unlabeled (PU) learning techniques have been employed by more and more financial institutions. However, most of these methods are designed for single data sets and do not take into account the heterogeneity of data when they are collected from different sources. In this paper, we propose an integrative PU learning method (I-PU) for pooling information from multiple heterogeneous PU data sets. A novel approach that penalizes group differences is developed to explicitly and automatically identify the cluster structures of coefficients across different data sets, thus offering a plausible interpretation of heterogeneity. Furthermore, we apply a bilevel selection method to detect the sparse structure at both the group level and within-group level. Theoretically, we show that our proposed estimator has the oracle property. Computationally, we design an expectation-maximization (EM) algorithm framework and propose an alternating direction method of multipliers (ADMM) algorithm to solve it. Simulation results show that our proposed method has better numerical performance in terms of variable selection, parameter estimation, and prediction ability. Finally, a real-world application showcases the effectiveness of our method in identifying distinct coefficient clusters and its superior prediction performance compared with direct data merging or separate modeling. This result also offers valuable insights for financial institutions in developing targeted fraud detection systems. History: Accepted by Ram Ramesh, Area Editor for Data Science & Machine Learning. Funding: This work was supported by the National Natural Science Foundation of China [Grants 72071169, 72231005, 72233002, and 72471169], the Fundamental Research Funds for the Central Universities of China [Grant 20720231060], the National Social Science Fund of China [Grant 21&ZD146], and Shuimu Tsinghua Scholar Program. Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2023.0366 ) as well as from the IJOC GitHub software repository ( https://github.com/INFORMSJoC/2023.0366 ). The complete IJOC Software and Data Repository is available at https://informsjoc.github.io/ .

The acceptance of academic papers involves a complex peer-review process that requires substantial human and material resources and is susceptible to biases. With advancements in deep learning technologies, researchers have explored automated approaches for assessing paper acceptance. Existing automated academic paper rating methods primarily rely on the full content of papers to estimate acceptance probabilities. However, these methods are often inefficient and introduce redundant or irrelevant information. Additionally, while Bert can capture general semantic representations through pretraining on large-scale corpora, its performance on the automatic academic paper rating (AAPR) task remains suboptimal due to discrepancies between its pretraining corpus and academic texts. To address these issues, this study proposes LMCBert, a model that integrates large language models (LLMs) with momentum contrastive learning (MoCo). LMCBert utilizes LLMs to extract the core semantic content of papers, reducing redundancy and improving the understanding of academic texts. Furthermore, it incorporates MoCo to optimize Bert training, enhancing the differentiation of semantic representations and improving the accuracy of paper acceptance predictions. Empirical evaluations demonstrate that LMCBert achieves effective performance on the evaluation dataset, supporting the validity of the proposed approach. The code and data used in this article are publicly available at https://github.com/iioSnail/LMCBert.

In the current wave of technological innovation, the scientific and technological evaluation system not only plays a crucial role in assessing scientific research achievements but also serves as a driving force for advancing scientific research. To enhance the effectiveness and accuracy of peer review in the scientific and technological evaluation system, an integrating fuzzy C-Means clustering algorithm and similarity analysis method is proposed. In this methodology, the fuzzy C-Means clustering algorithm is first applied to conduct cluster analysis on expert feature attribute. Second, the recursive feature elimination feature selection is used to select the important features for enhancing the model generalization capability. The Gower distance of key features between experts is calculated to determine the similarity between experts, and the association features among experts are mined. Finally, an expert credibility prediction model is constructed based on a feedforward neural network. A real-world expert credibility dataset was used to validate the effectiveness and feasibility of the proposed method. The experimental results indicated that integrating the fuzzy C-Means clustering algorithm and similarity association is an effective way to improve the model prediction performance. This can be considered a promising solution for the scientific and technological evaluation.

Anchor-based methods are proposed to make use of anchors to produce an affinity matrix of objects to improve the scalability of traditional spectral clustering (SC). Nevertheless, the membership heterogeneity of objects inside a cluster, which would bring about low quality of anchors and hurt the clustering accuracy, is commonly neglected by existing anchor-based algorithms. To address this problem, this article proposes a novel approach to adopt the Gaussian mixture model (GMM) to enhance anchor-based SC for large-scale data in a two-stage divide-and-conquer manner. In the first stage, GMM with expectation maximization (EM) algorithm is employed to divide the objects into two categories as prior-consistent objects and prior-uncertain objects in considering the membership heterogeneity of objects. In the second stage, anchor-based SC is conducted on the prior-uncertain objects by sampling the anchors from the Gaussian components derived from the first stage. Then, the produced clusters in the second stage are aligned with those Gaussian components by maximizing the membership of objects with respect to clusters. The computation complexity of the proposed GMM-SC approach is much smaller than that of the anchor-based SC. The experiments on large-scale datasets also validate the superiority of the proposed GMM-SC approach over state-of-the-art techniques.

The security of credit card fraud detection (CCFD) models based on machine learning is important but rarely considered in the existing research. To this end, we propose a black-box attack-based security evaluation framework for CCFD models. Under this framework, the semisupervised learning technique and transfer-based black-box attack are combined to construct two versions of a semisupervised transfer black-box attack algorithm. Moreover, we introduce a new nonlinear optimization model to generate the adversarial examples against CCFD models and a security evaluation index to quantitatively evaluate the security of them. Computing experiments on two real data sets demonstrate that, facing the adversarial examples generated by the proposed attack algorithms, all six supervised models considered largely lose their ability to identify the fraudulent transactions, whereas the two unsupervised models are less affected. This indicates that the CCFD models based on supervised machine learning may possess substantial security risks. In addition, the evaluation results for the security of the models generate important managerial implications that help banks reasonably evaluate and enhance the model security. History: Accepted by Ram Ramesh, Area Editor for Data Science & Machine Learning. Funding: This work was supported in part by the National Natural Science Foundation of China [Grants 72171160 and 71988101], Key Program of National Natural Science Foundation of China and Quebec Research Foundation (NSFC-FRQ) Joint Project [Grant 7191101304], Key Program of NSFC-FRQSC Joint Project [Grant 72061127002], Excellent Youth Foundation of Sichuan Province [Grant 2020JDJQ0021], and National Leading Talent Cultivation Project of Sichuan University [Grant SKSYL2021-03]. Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2023.1297 ) as well as from the IJOC GitHub software repository ( https://github.com/INFORMSJoC/2021.0076 ) at ( http://dx.doi.org/10.5281/zenodo.7631457 ).

Obtaining a multiresource allocation scheme for multitask influenced by uncertain factors is a critical problem in a collaborative logistics network. This paper presents an optimal allocation model of fuzzy resources for multistage random logistics tasks based on the six-point trapezoidal fuzzy number and the membership function. Besides, considering task demands and resource constraints, a new cost-time-quality multiobjective programming of N-N task-resource assignment is introduced, which can be divided into minimize total logistics cost and execution time, maximize total service quality. Furthermore, by setting the different simulation scenarios, the results show that if the decision maker has a higher risk preference and pursues the optimization of single or multiobjective, the higher degree membership and satisfaction function values can be obtained with a larger compensation coefficient. The allocation scheme of task-resource assignment generated by proposed model has a high global level of utilization efficiency, which can effectively utilize fuzzy resources in collaborative logistics network, and avoid resource shortage caused by the excessive occupation of local resources.

In this paper, some privacy-preserving features for distributed subgradient optimization algorithms are considered. Most of the existing distributed algorithms focus mainly on the algorithm design and convergence analysis, but not the protection of agents' privacy. Privacy is becoming an increasingly important issue in applications involving sensitive information. In this paper, we first show that the distributed subgradient synchronous homogeneous-stepsize algorithm is not privacy preserving in the sense that the malicious agent can asymptotically discover other agents' subgradients by transmitting untrue estimates to its neighbors. Then a distributed subgradient asynchronous heterogeneous-stepsize projection algorithm is proposed and accordingly its convergence and optimality is established. In contrast to the synchronous homogeneous-stepsize algorithm, in the new algorithm agents make their optimization updates asynchronously with heterogeneous stepsizes. The introduced two mechanisms of projection operation and asynchronous heterogeneous-stepsize optimization can guarantee that agents' privacy can be effectively protected.

A stock selection model with both discrete and continuous decision variables is proposed, in which a novel sigmoid-based mixed discrete-continuous differential evolution algorithm is especially developed for model optimization. In particular, a stock scoring mechanism is first designed to evaluate candidate stocks based on their fundamental and technical features, and the top-ranked stocks are selected to formulate an equal-weighted portfolio. Generally, the proposed model makes literature contributions from two main perspectives. First, to determine the optimal solution in terms of feature selections (discrete variables) and the corresponding weights (continuous variables), the original differential evolution algorithm focusing only on continuous problems is extended to a novel mixed discrete-continuous variant based on sigmoid-based conversion for the discrete part. Second, the stock selection model also resolves the gap of the application of differential evolution algorithm to stock selection. Using the Shanghai A share market of China as the study sample, the empirical results show that the novel stock selection model can make a profitable portfolio and significantly outperform its benchmarks (with other model designs and optimization algorithms used in the existing studies) in terms of both investment return and model robustness.

With the development of social economy, the increasing demand on hazardous materials transportation has received considerable attention around the world and the location-routing problem has become a hot topic of study. The goal of our work is to obtain the best balance between the transportation risk and cost under the assumptions that the transportation costs and the number of affected people are fuzzy variables. Within the framework of credibility theory, we propose a chance-constrained programming model and design a fuzzy simulation–based genetic algorithm to solve the model. Numerical experiments are given to illustrate the efficiency of the proposed model and algorithm.

A fuzzy number is a normal and convex fuzzy subset of the real line. In this paper, based on membership function, we redefine the concepts of mean and variance for fuzzy numbers. Furthermore, we propose the concept of skewness and prove some desirable properties. A fuzzy mean-variance-skewness portfolio selection model is formulated and two variations are given, which are transformed to nonlinear optimization models with polynomial objective and constraint functions such that they can be solved analytically. Finally, we present some numerical examples to demonstrate the effectiveness of the proposed models.

In this paper, an evolving least squares support vector machine (LSSVM) learning paradigm with a mixed kernel is proposed to explore stock market trends. In the proposed learning paradigm, a genetic algorithm (GA), one of the most popular evolutionary algorithms (EAs), is first used to select input features for LSSVM learning, i.e., evolution of input features. Then, another GA is used for parameters optimization of LSSVM, i.e., evolution of algorithmic parameters. Finally, the evolving LSSVM learning paradigm with best feature subset, optimal parameters, and a mixed kernel is used to predict stock market movement direction in terms of historical data series. For illustration and evaluation purposes, three important stock indices, S&P 500 Index, Dow Jones Industrial Average (DJIA) Index, and New York Stock Exchange (NYSE) Index, are used as testing targets. Experimental results obtained reveal that the proposed evolving LSSVM can produce some forecasting models that are easier to be interpreted by using a small number of predictive features and are more efficient than other parameter optimization methods. Furthermore, the produced forecasting model can significantly outperform other forecasting models listed in this paper in terms of the hit ratio. These findings imply that the proposed evolving LSSVM learning paradigm can be used as a promising approach to stock market tendency exploration.

In this study, a generalized Intelligent-agent-based fuzzy group forecasting model is proposed for oil price prediction. In the proposed model, some single Intelligent-agent-based predictors with much disagreement are first created for crude oil price prediction. Then these single prediction results produced by these single intelligent predictors are fuzzified into some fuzzy prediction representations. Particularly, some methods of fuzzification are extended into a consolidated framework to make the later computation generalization. Subsequently, these fuzzified prediction representations are integrated into a fuzzy consensus, i.e., aggregated fuzzy prediction. Finally, the aggregated fuzzy prediction is defuzzified into a crisp value as the final prediction results. For verification and testing purposes, two typical oil price series are used to conduct the experiments.

An effective foreign exchange (forex) trading decision is usually dependent on effective forex forecasting. In this study, an intelligent system framework integrating forex forecasting and trading decision is first proposed. Based on this framework, an advanced intelligent decision support system (DSS) incorporating a back-propagation neural network (BPNN)-based forex forecasting subsystem and Web-based forex trading decision support subsystem is developed, which has been used to predict the directional change of daily forex rates and provide intelligent online decision support for financial institutions and individual investors. This article describes the forex forecasting and trading decision method, the system architecture, main functions, and operation of the developed DSS system. A comparative study is conducted between our developed system and others commonly used in order to assess the overall performance of the developed system. The assessment results show that our developed DSS outperforms some commonly used forex forecasting and trading decision systems and can provide intelligent e-service for forex traders to make useful trading decisions in the forex market. © 2007 Wiley Periodicals, Inc. Int J Int Syst 22: 475–499, 2007.

In this study, we propose a double-phase neural-network-based metalearning approach to perform distributed text information retrieval. In the first phase, a single neural network model is deployed in different text collections distributed in different physical sites to retrieve some relevant text documents. In the second phase, a neural-network-based metalearning approach is proposed to integrate the relevance results for text documents with a specific query. For illustration purpose, a simulated web text information retrieval experiment is performed to verify the effectiveness and efficiency of the proposed neural-network-based metalearning approach.

Data preparation is an important and critical step in neural network modeling for complex data analysis and it has a huge impact on the success of a wide variety of complex data analysis tasks, such as data mining and knowledge discovery. Although data preparation in neural network data analysis is important, some existing literature about the neural network data preparation are scattered, and there is no systematic study about data preparation for neural network data analysis. In this study, we first propose an integrated data preparation scheme as a systematic study for neural network data analysis. In the integrated scheme, a survey of data preparation, focusing on problems with the data and corresponding processing techniques, is then provided. Meantime, some intelligent data preparation solution to some important issues and dilemmas with the integrated scheme are discussed in detail. Subsequently, a cost-benefit analysis framework for this integrated scheme is presented to analyze the effect of data preparation on complex data analysis. Finally, a typical example of complex data analysis from the financial domain is provided in order to show the application of data preparation techniques and to demonstrate the impact of data preparation on complex data analysis.

In this study, a hybrid artificial intelligent (AI) system integrating neural network and expert system is proposed to support foreign exchange (forex) trading decisions. In this system, a neural network is used to predict the forex price in terms of quantitative data, while an expert system is used to handle qualitative factor and to provide forex trading decision suggestions for traders incorporating experts' knowledge and the neural network's results. The effectiveness of the proposed hybrid AI system is illustrated by simulation experiments.