Decision Support Systems最新文献_第9页

Handling imperfection: A taxonomy for machine learning on data with data quality defects 处理缺陷：一种针对具有数据质量缺陷的数据进行机器学习的分类法

IF 6.7 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Decision Support Systems

Pub Date : 2025-06-16 DOI: 10.1016/j.dss.2025.114493

Michael Hagn, Bernd Heinrich, Thomas Krapf, Alexander Schiller

In recent years, machine learning (ML) has become ubiquitous in sectors including transportation, security, health, and finance to analyze large amounts of data and support decision-making. However, real-world datasets used in ML often exhibit various data quality (DQ) defects that can significantly impair the performance and validity of ML models and thus also the decisions derived from them. Therefore, a plethora of methods across various research strands have been proposed to address DQ defects and mitigate their negative impact on ML-based data analysis and decision support. This has resulted in a fragmented research landscape, where comparisons and classifications of methods dealing with ML on data with DQ defects are very challenging for both researchers and practitioners. Thus, based on a structured design process, we develop and present a taxonomy for this research field. The taxonomy serves as a systematic framework to classify and organize existing research and methods according to relevant dimensions and facilitates future work in this area. Its reliability, understandability, completeness, and usefulness are supported by an evaluation with external researchers and practitioners. Finally, we identify current trends and research gaps and derive challenges and directions for future research.

近年来，机器学习（ML）在交通、安全、健康和金融等领域无处不在，可以分析大量数据并支持决策。然而，机器学习中使用的真实数据集经常表现出各种数据质量（DQ）缺陷，这些缺陷会严重损害机器学习模型的性能和有效性，从而也会损害从中得出的决策。因此，已经提出了跨越各种研究领域的大量方法来解决DQ缺陷，并减轻它们对基于ml的数据分析和决策支持的负面影响。这导致了一个支离破碎的研究领域，其中比较和分类的方法处理ML的数据与DQ缺陷是非常具有挑战性的研究人员和从业者。因此，基于一个结构化的设计过程，我们为这个研究领域开发并提出了一个分类法。该分类法作为一个系统的框架，将现有的研究和方法按照相关的维度进行分类和组织，并有助于今后在这一领域的工作。它的可靠性，可理解性，完整性和有用性是由外部研究人员和实践者的评估支持的。最后，我们确定了当前的趋势和研究差距，并得出了未来研究的挑战和方向。

{"title":"Handling imperfection: A taxonomy for machine learning on data with data quality defects","authors":"Michael Hagn, Bernd Heinrich, Thomas Krapf, Alexander Schiller","doi":"10.1016/j.dss.2025.114493","DOIUrl":"10.1016/j.dss.2025.114493","url":null,"abstract":"<div><div>In recent years, machine learning (ML) has become ubiquitous in sectors including transportation, security, health, and finance to analyze large amounts of data and support decision-making. However, real-world datasets used in ML often exhibit various data quality (DQ) defects that can significantly impair the performance and validity of ML models and thus also the decisions derived from them. Therefore, a plethora of methods across various research strands have been proposed to address DQ defects and mitigate their negative impact on ML-based data analysis and decision support. This has resulted in a fragmented research landscape, where comparisons and classifications of methods dealing with ML on data with DQ defects are very challenging for both researchers and practitioners. Thus, based on a structured design process, we develop and present a taxonomy for this research field. The taxonomy serves as a systematic framework to classify and organize existing research and methods according to relevant dimensions and facilitates future work in this area. Its reliability, understandability, completeness, and usefulness are supported by an evaluation with external researchers and practitioners. Finally, we identify current trends and research gaps and derive challenges and directions for future research.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"196 ","pages":"Article 114493"},"PeriodicalIF":6.7,"publicationDate":"2025-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144338972","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

An explainable lesion detection transformer model for medical imaging diagnosis decision support: Design science research 用于医学影像诊断决策支持的可解释病变检测变压器模型：设计科学研究

IF 6.7 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Decision Support Systems

Pub Date : 2025-06-16 DOI: 10.1016/j.dss.2025.114492

Xinwei Wang , Yi Feng , Sutong Wang , Dujuan Wang , T.C.E. Cheng

Utilizing machine learning methods for auxiliary decision support in medical imaging significantly reduces missed detections and unnecessary expenses. However, the strict accuracy and transparency requirements in the medical field pose challenges for deep learning applications based on neural networks. To address these issues, we propose a novel artificial intelligence artifact guided by the design science research methodology for lesion detection decision support in medical images, called Explainable Lesion DEtection TRansformer (EL-DETR). This approach features an explainable separate attention mechanism in the decoder that highlights the attention weights of content and location queries, providing insights into the inference process through attention mapping visualizations. In addition, we introduce a hybrid matching query strategy to enhance the learning of positive samples and develop an adaptive efficient compound loss function to optimize training. We demonstrate EL-DETR's superior accuracy, robustness, and interpretability using four real-world datasets, establishing it as a reliable tool for clinical diagnosis and treatment decision support based on medical imaging. The code and original data are available at https://github.com/weimingai/EL-DETR.

利用机器学习方法在医学成像中进行辅助决策支持，可以显著减少漏检和不必要的费用。然而，医学领域对准确性和透明度的严格要求给基于神经网络的深度学习应用带来了挑战。为了解决这些问题，我们提出了一种新的人工智能工件，以设计科学研究方法为指导，用于医学图像中的病变检测决策支持，称为可解释病变检测变压器（EL-DETR）。这种方法在解码器中具有可解释的独立注意机制，突出显示内容和位置查询的注意权重，通过注意映射可视化提供对推理过程的见解。此外，我们引入了一种混合匹配查询策略来增强正样本的学习，并开发了一种自适应的高效复合损失函数来优化训练。我们利用四个真实世界的数据集证明了EL-DETR优越的准确性、稳健性和可解释性，并将其建立为基于医学成像的临床诊断和治疗决策支持的可靠工具。代码和原始数据可在https://github.com/weimingai/EL-DETR上获得。

{"title":"An explainable lesion detection transformer model for medical imaging diagnosis decision support: Design science research","authors":"Xinwei Wang , Yi Feng , Sutong Wang , Dujuan Wang , T.C.E. Cheng","doi":"10.1016/j.dss.2025.114492","DOIUrl":"10.1016/j.dss.2025.114492","url":null,"abstract":"<div><div>Utilizing machine learning methods for auxiliary decision support in medical imaging significantly reduces missed detections and unnecessary expenses. However, the strict accuracy and transparency requirements in the medical field pose challenges for deep learning applications based on neural networks. To address these issues, we propose a novel artificial intelligence artifact guided by the design science research methodology for lesion detection decision support in medical images, called Explainable Lesion DEtection TRansformer (EL-DETR). This approach features an explainable separate attention mechanism in the decoder that highlights the attention weights of content and location queries, providing insights into the inference process through attention mapping visualizations. In addition, we introduce a hybrid matching query strategy to enhance the learning of positive samples and develop an adaptive efficient compound loss function to optimize training. We demonstrate EL-DETR's superior accuracy, robustness, and interpretability using four real-world datasets, establishing it as a reliable tool for clinical diagnosis and treatment decision support based on medical imaging. The code and original data are available at <span><span>https://github.com/weimingai/EL-DETR</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"196 ","pages":"Article 114492"},"PeriodicalIF":6.7,"publicationDate":"2025-06-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144291619","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Investigating the impact of differential privacy obfuscation on users’ data disclosure decisions 调查不同隐私混淆对用户数据披露决策的影响

IF 6.7 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Decision Support Systems

Pub Date : 2025-06-03 DOI: 10.1016/j.dss.2025.114474

Michael Khavkin, Eran Toch

Differential Privacy (DP) has emerged as the standard for privacy-preserving analysis of individual-level data. Despite growing attention in the research community to the operationalization of DP through the selection of the privacy budget

ɛ

, little is known about how DP obfuscation affects users’ disclosure decisions in data market scenarios. These decisions may be context-specific and vary with privacy preferences, eliciting disparate data valuations across individuals. Through a choice-based conjoint analysis (

N_{1} = 588

), simulating realistic data markets, we analyzed how varying DP protection levels influence individual decision-making of participation in data collection under DP. Our findings show that personal reward and the guaranteed DP protection had the strongest influence on participants’ selection of a data collection scenario. Surprisingly, the type of disclosed data had the least influence on participants’ decisions to disclose personal data, a trend consistent across participants from different countries. Furthermore, increasing the DP protection level by a single unit reduced the preferred compensation price by over 60% for the same level of user utility, with marginal effects diminishing exponentially at higher DP levels. Our results were then confirmed in an online study (

N_{2} = 146

) involving real data disclosure with actual payments, using our original scenario framing. Our findings can support context-specific DP configuration and help data practitioners improve decision-making associated with privacy protection in differentially private systems, balancing the trade-off between DP and compensation costs.

差分隐私（DP）已成为个人数据隐私保护分析的标准。尽管研究界越来越关注通过选择隐私预算来实现DP的操作化，但对于数据市场场景下DP混淆如何影响用户的披露决策却知之甚少。这些决定可能是特定于环境的，并且随着隐私偏好的变化而变化，从而引起个人之间不同的数据估值。通过基于选择的联合分析（N1=588），模拟现实数据市场，我们分析了不同的数据保护水平如何影响个人参与数据收集的决策。我们的研究结果表明，个人奖励和保证DP保护对参与者选择数据收集场景的影响最大。令人惊讶的是，披露的数据类型对参与者披露个人数据的决定影响最小，这一趋势在不同国家的参与者中是一致的。此外，在相同的用户效用水平下，每增加一个单位的DP保护水平，可使首选补偿价格降低60%以上，而在更高的DP水平下，边际效应呈指数级递减。我们的结果随后在一项在线研究（N2=146）中得到证实，该研究涉及实际支付的真实数据披露，使用我们的原始场景框架。我们的研究结果可以支持上下文特定的DP配置，并帮助数据从业者在不同的私有系统中改进与隐私保护相关的决策，平衡DP和补偿成本之间的权衡。

{"title":"Investigating the impact of differential privacy obfuscation on users’ data disclosure decisions","authors":"Michael Khavkin, Eran Toch","doi":"10.1016/j.dss.2025.114474","DOIUrl":"10.1016/j.dss.2025.114474","url":null,"abstract":"<div><div>Differential Privacy (DP) has emerged as the standard for privacy-preserving analysis of individual-level data. Despite growing attention in the research community to the operationalization of DP through the selection of the privacy budget <span><math><mi>ɛ</mi></math></span>, little is known about how DP obfuscation affects users’ disclosure decisions in data market scenarios. These decisions may be context-specific and vary with privacy preferences, eliciting disparate data valuations across individuals. Through a choice-based conjoint analysis (<span><math><mrow><msub><mrow><mi>N</mi></mrow><mrow><mn>1</mn></mrow></msub><mo>=</mo><mn>588</mn></mrow></math></span>), simulating realistic data markets, we analyzed how varying DP protection levels influence individual decision-making of participation in data collection under DP. Our findings show that personal reward and the guaranteed DP protection had the strongest influence on participants’ selection of a data collection scenario. Surprisingly, the type of disclosed data had the least influence on participants’ decisions to disclose personal data, a trend consistent across participants from different countries. Furthermore, increasing the DP protection level by a single unit reduced the preferred compensation price by over 60% for the same level of user utility, with marginal effects diminishing exponentially at higher DP levels. Our results were then confirmed in an online study (<span><math><mrow><msub><mrow><mi>N</mi></mrow><mrow><mn>2</mn></mrow></msub><mo>=</mo><mn>146</mn></mrow></math></span>) involving real data disclosure with actual payments, using our original scenario framing. Our findings can support context-specific DP configuration and help data practitioners improve decision-making associated with privacy protection in differentially private systems, balancing the trade-off between DP and compensation costs.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"196 ","pages":"Article 114474"},"PeriodicalIF":6.7,"publicationDate":"2025-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144254153","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Social media meets FinTech platforms: How do online emotions support credit risk decision-making? 社交媒体与金融科技平台：网络情绪如何支持信用风险决策？

IF 6.7 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Decision Support Systems

Pub Date : 2025-05-27 DOI: 10.1016/j.dss.2025.114471

Zenan Zhou , Zhichen Chen , Yingjie Zhang , Tian Lu , Xianghua Lu

As emerging FinTech platforms face pressure in efficiently managing credit risk, the human emotional spectrum of FinTech platform borrowers within social media becomes a potential source for gaining insight into and evaluating their financial behaviors. Collaborating with an Asian FinTech platform, we investigate the impact of social media emotions on a platform’s loan-approval decisions and repayment-reminder interventions before due dates. We demonstrate that anger at the pre-approval stage has a U-shaped relationship with platform borrowers’ default probability. We reveal what we call “a bright side of anger” with respect to curbing financial credit risk: moderate intensity of anger at the pre-approval stage suggests a lower loan default probability. We also find that the average happiness tendency of platform delinquent borrowers’ at the pre-maturity stage becomes informative and valuable, as it shows a U-shaped relationship with loan default; as for anger, it does not work therein. Furthermore, our field experiment indicates that a positive-expectation reminder is useful for prompting repayment when delinquent borrowers are in strong emotional intensities, regardless of anger or happiness. However, a negative-consequence reminder results in a higher default probability for delinquent borrowers who maintain high immediate happiness before the loan maturity dates. We draw on the classical appraisal theory of emotions and the feelings-as-information theory to interpret our findings. We offer non-trivial theoretical and practical implications to support FinTech platform credit risk decision-making by investigating the value of social media emotions and advocating for cross-functional coordination between debt approval and debt collection departments.

随着新兴金融科技平台在有效管理信用风险方面面临压力，社交媒体上金融科技平台借款人的人类情感谱成为了解和评估其金融行为的潜在来源。我们与一家亚洲金融科技平台合作，调查了社交媒体情绪对平台贷款审批决策和到期前还款提醒干预的影响。我们证明了预审批阶段的愤怒与平台借款人的违约概率呈u型关系。在抑制金融信贷风险方面，我们揭示了我们所谓的“愤怒的光明面”：在审批前阶段，适度的愤怒意味着较低的贷款违约概率。我们还发现，平台违约借款人在前期的平均幸福倾向与贷款违约呈u型关系，具有信息价值；至于愤怒，它在那里不起作用。此外，我们的实地实验表明，当拖欠借款人处于强烈的情绪强度时，无论是愤怒还是快乐，积极期望提醒都有助于促使还款。然而，对于在贷款到期日之前保持高即时幸福感的违约借款人来说，负后果提醒会导致更高的违约概率。我们利用经典的情绪评价理论和感觉作为信息理论来解释我们的发现。我们通过调查社交媒体情绪的价值，并倡导债务审批和债务催收部门之间的跨职能协调，为支持金融科技平台的信用风险决策提供了重要的理论和实践意义。

{"title":"Social media meets FinTech platforms: How do online emotions support credit risk decision-making?","authors":"Zenan Zhou , Zhichen Chen , Yingjie Zhang , Tian Lu , Xianghua Lu","doi":"10.1016/j.dss.2025.114471","DOIUrl":"10.1016/j.dss.2025.114471","url":null,"abstract":"<div><div>As emerging FinTech platforms face pressure in efficiently managing credit risk, the human emotional spectrum of FinTech platform borrowers within social media becomes a potential source for gaining insight into and evaluating their financial behaviors. Collaborating with an Asian FinTech platform, we investigate the impact of social media emotions on a platform’s loan-approval decisions and repayment-reminder interventions before due dates. We demonstrate that anger at the pre-approval stage has a U-shaped relationship with platform borrowers’ default probability. We reveal what we call “<em>a bright side of anger</em>” with respect to curbing financial credit risk: moderate intensity of anger at the pre-approval stage suggests a lower loan default probability. We also find that the average happiness tendency of platform delinquent borrowers’ at the pre-maturity stage becomes informative and valuable, as it shows a U-shaped relationship with loan default; as for anger, it does not work therein. Furthermore, our field experiment indicates that a positive-expectation reminder is useful for prompting repayment when delinquent borrowers are in strong emotional intensities, regardless of anger or happiness. However, a negative-consequence reminder results in a higher default probability for delinquent borrowers who maintain high immediate happiness before the loan maturity dates. We draw on the classical appraisal theory of emotions and the feelings-as-information theory to interpret our findings. We offer non-trivial theoretical and practical implications to support FinTech platform credit risk decision-making by investigating the value of social media emotions and advocating for cross-functional coordination between debt approval and debt collection departments.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"195 ","pages":"Article 114471"},"PeriodicalIF":6.7,"publicationDate":"2025-05-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144167967","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Multi-interest transfer using contrastive learning for cross-domain recommendation 基于对比学习的多兴趣迁移跨领域推荐

IF 6.7 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Decision Support Systems

Pub Date : 2025-05-23 DOI: 10.1016/j.dss.2025.114473

Yu-Lin Lai , Szu-Hao Huang , Chiao-Ting Chen , Cheng-Jhang Wu

With advancements in information technology, deep learning techniques have been widely applied to recommendation systems, substantially assisting businesses and users in making better decisions. However, it still faces some intractable limitations, such as the cold-start problem and data sparsity. Hence, cross-domain recommendations are proposed to address these problems by referring to the domains with richer data. Existing models usually apply domain- or user-level transferal to exchange information between domains. For domain-level transferals, information is transferred directly using a straightforward transformation without filtering. In contrast, user-level transferal sets trainable parameters to control the ratio of user embedding from two domains. The former is insufficiently precise for every user, and the latter encounters generalization issues. For these reasons, these methods ameliorate the cold-start problem but create a new problem: negative transfer. Thus, we propose an interest-level transferal called multi-interest transferal to more precisely extract multiple interests and transfer related ones according to the target items. Nevertheless, it is not easy to model interest correlations of different domains. We, therefore, devise three self-supervised learning tasks to model the correlations and extract discriminant information. The experimental results reveal that this model outperforms other state-of-the-art methods by about 7% to 10%. Through multi-interest and contrastive learning techniques, our approach can model the decision-making process more effectively in cross-domain recommendation.

随着信息技术的进步，深度学习技术已被广泛应用于推荐系统，极大地帮助企业和用户做出更好的决策。然而，它仍然面临一些棘手的限制，如冷启动问题和数据稀疏性。因此，为了解决这些问题，我们提出了跨领域的建议，即参考具有更丰富数据的领域。现有模型通常采用域级或用户级转移来交换域之间的信息。对于域级传输，信息直接使用直接转换而不进行过滤。相比之下，用户级转移设置可训练的参数来控制用户从两个域嵌入的比例。前者对每个用户都不够精确，而后者则遇到泛化问题。由于这些原因，这些方法改善了冷启动问题，但产生了一个新的问题：负转移。因此，我们提出了一种利益层面的转移，称为多利益转移，可以更精确地提取多个利益，并根据目标项目转移相关的利益。然而，对不同领域的兴趣相关性进行建模并不容易。因此，我们设计了三个自监督学习任务来建模相关性并提取判别信息。实验结果表明，该模型的性能优于其他最先进的方法约7%至10%。通过多兴趣和对比学习技术，我们的方法可以更有效地模拟跨领域推荐的决策过程。

{"title":"Multi-interest transfer using contrastive learning for cross-domain recommendation","authors":"Yu-Lin Lai , Szu-Hao Huang , Chiao-Ting Chen , Cheng-Jhang Wu","doi":"10.1016/j.dss.2025.114473","DOIUrl":"10.1016/j.dss.2025.114473","url":null,"abstract":"<div><div>With advancements in information technology, deep learning techniques have been widely applied to recommendation systems, substantially assisting businesses and users in making better decisions. However, it still faces some intractable limitations, such as the cold-start problem and data sparsity. Hence, cross-domain recommendations are proposed to address these problems by referring to the domains with richer data. Existing models usually apply domain- or user-level transferal to exchange information between domains. For domain-level transferals, information is transferred directly using a straightforward transformation without filtering. In contrast, user-level transferal sets trainable parameters to control the ratio of user embedding from two domains. The former is insufficiently precise for every user, and the latter encounters generalization issues. For these reasons, these methods ameliorate the cold-start problem but create a new problem: negative transfer. Thus, we propose an interest-level transferal called multi-interest transferal to more precisely extract multiple interests and transfer related ones according to the target items. Nevertheless, it is not easy to model interest correlations of different domains. We, therefore, devise three self-supervised learning tasks to model the correlations and extract discriminant information. The experimental results reveal that this model outperforms other state-of-the-art methods by about 7% to 10%. Through multi-interest and contrastive learning techniques, our approach can model the decision-making process more effectively in cross-domain recommendation.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"195 ","pages":"Article 114473"},"PeriodicalIF":6.7,"publicationDate":"2025-05-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144185389","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

More than meets the eye: Feature concerns and suggestions in mobile XR app reviews 手机XR应用评论中的功能关注点和建议

IF 6.7 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Decision Support Systems

Pub Date : 2025-05-19 DOI: 10.1016/j.dss.2025.114475

Nohel Zaman , David M. Goldberg , Zhilei Qiao , Alan S. Abrahams

This study aims to address quality concerns in mobile Extended Reality (XR) apps by developing tools to classify user reviews from the Google Play Store. Our first major contribution is a holistic approach to coding thousands of reviews, offering a unified framework to assess feature concerns and feature suggestions, thereby enhancing the understanding of user expectations. We also introduce a novel toolset for extracting key terms related to these concerns and suggestions, facilitating rapid analysis and improvement of app functionalities. To validate our approach, we compare the performance of this automated coding method with sentiment analysis, deep learning-based BERT, and several large language models (LLMs). Additionally, our study extends the understanding of user satisfaction by analyzing quantitative indicators such as app ratings and review helpfulness, offering new perspectives on user perceptions of XR app quality. This research has discovered vital areas for feature improvement without implementing complex infrastructure and using advanced technical expertise, enabling app firms to better target their efforts in improving XR app experiences.

本研究旨在通过开发工具对b谷歌Play Store的用户评论进行分类，解决移动扩展现实（XR）应用的质量问题。我们的第一个主要贡献是对数千个评论进行编码的整体方法，提供了一个统一的框架来评估功能关注点和功能建议，从而增强了对用户期望的理解。我们还引入了一个新的工具集，用于提取与这些问题和建议相关的关键术语，促进快速分析和改进应用程序功能。为了验证我们的方法，我们将这种自动编码方法与情感分析、基于深度学习的BERT和几个大型语言模型（llm）的性能进行了比较。此外，我们的研究通过分析应用评级和评论帮助度等量化指标，扩展了对用户满意度的理解，为用户对XR应用质量的看法提供了新的视角。这项研究发现了在不实施复杂基础设施和使用先进技术专业知识的情况下进行功能改进的关键领域，使应用程序公司能够更好地瞄准改善XR应用程序体验的目标。

{"title":"More than meets the eye: Feature concerns and suggestions in mobile XR app reviews","authors":"Nohel Zaman , David M. Goldberg , Zhilei Qiao , Alan S. Abrahams","doi":"10.1016/j.dss.2025.114475","DOIUrl":"10.1016/j.dss.2025.114475","url":null,"abstract":"<div><div>This study aims to address quality concerns in mobile Extended Reality (XR) apps by developing tools to classify user reviews from the Google Play Store. Our first major contribution is a holistic approach to coding thousands of reviews, offering a unified framework to assess feature concerns and feature suggestions, thereby enhancing the understanding of user expectations. We also introduce a novel toolset for extracting key terms related to these concerns and suggestions, facilitating rapid analysis and improvement of app functionalities. To validate our approach, we compare the performance of this automated coding method with sentiment analysis, deep learning-based BERT, and several large language models (LLMs). Additionally, our study extends the understanding of user satisfaction by analyzing quantitative indicators such as app ratings and review helpfulness, offering new perspectives on user perceptions of XR app quality. This research has discovered vital areas for feature improvement without implementing complex infrastructure and using advanced technical expertise, enabling app firms to better target their efforts in improving XR app experiences.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"195 ","pages":"Article 114475"},"PeriodicalIF":6.7,"publicationDate":"2025-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144098457","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Integrating temporal association rules into intelligent prediction system for metabolic dysfunction-associated fatty liver disease 将时间关联规则集成到代谢功能障碍相关脂肪肝疾病智能预测系统中

IF 6.7 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Decision Support Systems

Pub Date : 2025-05-13 DOI: 10.1016/j.dss.2025.114467

Zhuoqing Wu , Chonghui Guo , Jingfeng Chen , Suying Ding , Yunchao Zheng

Healthcare big data provides trajectory data on chronic disease onset, progression, and outcomes, essential for understanding metabolic dysfunction-associated fatty liver disease (MAFLD) patient health dynamics. However, constructing explainable predictive models for MAFLD using longitudinal healthcare big data remains challenging due to its complexity. While several high-performance machine learning models have shown promise, their “black box” nature limits interpretability and trust among clinical healthcare professionals. Most studies also rely on cross-sectional data, which lacks the depth of longitudinal data, hindering accurate health status tracking. This paper proposes an intelligent MAFLD prediction system integrating temporal association rules (TARs) through a “human-in-the-loop” approach. By analyzing TARs that capture disease dynamics, the system incorporates high-quality domain knowledge into its predictive model. To enhance explainability, we use the SHapley Additive exPlanations framework alongside clinically significant TARs. The system’s effectiveness was validated on real-world data, showing improved MAFLD outcome prediction. Sensitivity analysis identified optimal TARs and robust model configurations. Finally, the online-deployed explainable prototype system demonstrates potential to boost trust and adoption among clinical healthcare professionals. Additionally, the system’s effectiveness and their willingness to use it were further evaluated through the “human-on-the-loop” method. These findings suggest the system could serve as a valuable tool for clinical applications and advance information systems design.

医疗保健大数据提供慢性疾病发病、进展和结果的轨迹数据，对于了解代谢功能障碍相关脂肪肝（MAFLD）患者的健康动态至关重要。然而，由于其复杂性，使用纵向医疗保健大数据构建可解释的MAFLD预测模型仍然具有挑战性。虽然一些高性能机器学习模型显示出了希望，但它们的“黑箱”性质限制了临床医疗保健专业人员的可解释性和信任度。大多数研究还依赖于横断面数据，缺乏纵向数据的深度，阻碍了准确的健康状况跟踪。本文通过“人在环”的方法，提出了一种集成时间关联规则（TARs）的MAFLD智能预测系统。通过分析捕捉疾病动态的TARs，系统将高质量的领域知识整合到其预测模型中。为了提高可解释性，我们将SHapley加性解释框架与临床显著的tar一起使用。在实际数据中验证了该系统的有效性，显示出改进的MAFLD结果预测。灵敏度分析确定了最优TARs和鲁棒模型配置。最后，在线部署的可解释原型系统展示了在临床医疗保健专业人员中提高信任和采用的潜力。此外，通过“人在循环”方法进一步评估了系统的有效性和他们使用它的意愿。这些发现表明该系统可以作为临床应用和先进信息系统设计的有价值的工具。

{"title":"Integrating temporal association rules into intelligent prediction system for metabolic dysfunction-associated fatty liver disease","authors":"Zhuoqing Wu , Chonghui Guo , Jingfeng Chen , Suying Ding , Yunchao Zheng","doi":"10.1016/j.dss.2025.114467","DOIUrl":"10.1016/j.dss.2025.114467","url":null,"abstract":"<div><div>Healthcare big data provides trajectory data on chronic disease onset, progression, and outcomes, essential for understanding metabolic dysfunction-associated fatty liver disease (MAFLD) patient health dynamics. However, constructing explainable predictive models for MAFLD using longitudinal healthcare big data remains challenging due to its complexity. While several high-performance machine learning models have shown promise, their “black box” nature limits interpretability and trust among clinical healthcare professionals. Most studies also rely on cross-sectional data, which lacks the depth of longitudinal data, hindering accurate health status tracking. This paper proposes an intelligent MAFLD prediction system integrating temporal association rules (TARs) through a “human-in-the-loop” approach. By analyzing TARs that capture disease dynamics, the system incorporates high-quality domain knowledge into its predictive model. To enhance explainability, we use the SHapley Additive exPlanations framework alongside clinically significant TARs. The system’s effectiveness was validated on real-world data, showing improved MAFLD outcome prediction. Sensitivity analysis identified optimal TARs and robust model configurations. Finally, the online-deployed explainable prototype system demonstrates potential to boost trust and adoption among clinical healthcare professionals. Additionally, the system’s effectiveness and their willingness to use it were further evaluated through the “human-on-the-loop” method. These findings suggest the system could serve as a valuable tool for clinical applications and advance information systems design.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"195 ","pages":"Article 114467"},"PeriodicalIF":6.7,"publicationDate":"2025-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144084474","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

40 years of Decision Support Systems: A bibliometric analysis

IF 6.7 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Decision Support Systems

Pub Date : 2025-05-08 DOI: 10.1016/j.dss.2025.114469

Li Guan, José M. Merigó, Ghassan Beydoun

Decision Support Systems (DSS) is a leading international journal dedicated to decision support system research and practice, with the aim of exploring theoretical and technical advancements to facilitate enhanced decision making in industry, commerce, government, and other business settings. The journal published its first issue in 1985, and in 2025, celebrates its 40th anniversary. Motivated by this special event, this paper develops a comprehensive bibliometric analysis to present a lifetime overview of the development characteristics and leading trends of DSS journal between 1985 and 2023. By using the bibliographic data collected from the Scopus and Web of Science Core Collection databases, this study analyzes the publication and citation structure of the journal and investigates a wide range of issues including the most cited papers, the most cited documents by the journal's publications, the citing articles, the most productive and influential authors, institutions and countries/territories, and the most popular keywords and topics. Moreover, this work also graphically maps the bibliographic material by using the visualization of similarities (VOS) viewer software. In the graphical analysis, several bibliometric techniques in terms of co-citation, bibliographic coupling, and co-occurrence of author keywords are adopted. The results accentuate the significant growth and impact of DSS journal throughout its lifetime. It is expected that the journal will continue to grow its international reputation and disseminate knowledge in decision support, information systems, and business area, providing an efficient mechanism for researchers around the world to keep abreast with advances in the scientific community.

决策支持系统（DSS）是一本致力于决策支持系统研究和实践的国际领先期刊，旨在探索理论和技术进步，以促进工业，商业，政府和其他商业环境中的决策制定。该杂志于1985年出版了第一期，并于2025年庆祝其40周年。基于这一特殊事件，本文采用文献计量分析方法，对1985 - 2023年DSS期刊的发展特征和主要趋势进行了全面的回顾。本研究利用Scopus和Web of Science Core Collection数据库收集的文献数据，分析了该期刊的出版和被引结构，调查了该期刊被引次数最多的论文、被引次数最多的文献、被引文章、最高产和最具影响力的作者、机构和国家/地区、最热门的关键词和主题等问题。此外，本工作还使用相似度可视化（VOS）查看器软件对书目材料进行图形化映射。在图形分析中，采用了共被引、书目耦合和作者关键词共现等文献计量学技术。这些结果突出了DSS期刊在其整个生命周期中的显著增长和影响。预计该杂志将继续提高其国际声誉，并在决策支持、信息系统和商业领域传播知识，为世界各地的研究人员提供与科学界进展同步的有效机制。

{"title":"40 years of Decision Support Systems: A bibliometric analysis","authors":"Li Guan, José M. Merigó, Ghassan Beydoun","doi":"10.1016/j.dss.2025.114469","DOIUrl":"10.1016/j.dss.2025.114469","url":null,"abstract":"<div><div><em>Decision Support Systems</em> (DSS) is a leading international journal dedicated to decision support system research and practice, with the aim of exploring theoretical and technical advancements to facilitate enhanced decision making in industry, commerce, government, and other business settings. The journal published its first issue in 1985, and in 2025, celebrates its 40th anniversary. Motivated by this special event, this paper develops a comprehensive bibliometric analysis to present a lifetime overview of the development characteristics and leading trends of DSS journal between 1985 and 2023. By using the bibliographic data collected from the Scopus and Web of Science Core Collection databases, this study analyzes the publication and citation structure of the journal and investigates a wide range of issues including the most cited papers, the most cited documents by the journal's publications, the citing articles, the most productive and influential authors, institutions and countries/territories, and the most popular keywords and topics. Moreover, this work also graphically maps the bibliographic material by using the visualization of similarities (VOS) viewer software. In the graphical analysis, several bibliometric techniques in terms of co-citation, bibliographic coupling, and co-occurrence of author keywords are adopted. The results accentuate the significant growth and impact of DSS journal throughout its lifetime. It is expected that the journal will continue to grow its international reputation and disseminate knowledge in decision support, information systems, and business area, providing an efficient mechanism for researchers around the world to keep abreast with advances in the scientific community.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"194 ","pages":"Article 114469"},"PeriodicalIF":6.7,"publicationDate":"2025-05-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143943493","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Unlocking big data success in the AI-driven era: Toward a unified theory for intelligent decision support 在人工智能驱动时代解锁大数据的成功：迈向智能决策支持的统一理论

IF 6.7 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Decision Support Systems

Pub Date : 2025-05-08 DOI: 10.1016/j.dss.2025.114468

Xi Zhao , Hua Dai , Tao “Eric” Hu , Hsing K. Cheng , Ping Zhang

Upon a grounded theory-based literature review of 220 articles published in the AIS “Senior Scholars' Basket of Journals” over the period of twenty years of 2000–2020, this study examines concepts, constructs, topics, methodologies, and research models/paradigms of Big Data literature in the information systems (IS) discipline. We extend the well-established IS success model into the Big Data area, synthesize theoretical perspectives and empirical findings of literature, identify critical success factors and interrelationships, and develop a unified Big Data success theory in the organizational context. Building upon the literature review, we propose a set of research agendas and articulate opportunities and challenges of the evolving Big Data literature. The paper concludes with research implications, contributions, and limitations of the study in the ever-emerging AI-Driven era.

本研究对2000-2020年20年间发表在AIS“高级学者期刊篮子”上的220篇文章进行了基于理论的文献综述，考察了信息系统（IS）学科中大数据文献的概念、结构、主题、方法和研究模型/范式。我们将成熟的信息系统成功模型扩展到大数据领域，综合理论观点和文献的实证发现，识别关键成功因素和相互关系，并在组织环境中形成统一的大数据成功理论。在文献综述的基础上，我们提出了一套研究议程，并阐明了大数据文献发展的机遇和挑战。论文最后总结了研究的意义、贡献以及在不断涌现的人工智能驱动时代研究的局限性。

引用次数: 0

Product return prediction in live streaming e-commerce with cross-modal contrastive transformer 基于跨模态对比变压器的电商直播产品退货预测

IF 6.7 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Decision Support Systems

Pub Date : 2025-05-05 DOI: 10.1016/j.dss.2025.114470

Wen Zhang , Rui Xie , Pei Quan , Zhenzhong Ma

The live-streaming e-commerce industry is suffering heavy economic losses due to the high product return rate, which leads to rising logistics costs, greater inventory pressure, and unsatisfactory consumer experiences. Accurate product return prediction is highly desirable for the vendors to optimize their business operations in advance to reduce return-related costs. This paper proposes a novel approach, called Contraformer (Contrastive transformer), to predict product returns in live streaming e-commerce by leveraging fine-grained streamer behavior features extracted from three modalities (i.e., visual, acoustic, and language). The primary contribution lies in that we adopt Transformer with the encoder-decoder architecture with a novel class-supervised contrastive learning (CSCL) to fuse streamer behavior for multimodal representation alignment and inter-modal interaction characterization. By using a real-world dataset with 2584 product streamers and 864 items collected from Tiktok China live streaming platform, we demonstrate that the proposed Contrasformer approach outperforms the baseline methods in predicting product return rate with a 25 % reduction in terms of mean absolute error. This study offers great managerial implications for vendors to manage their practice in live streaming commerce.

由于产品退货率高，导致物流成本上升，库存压力加大，消费者体验不满意，直播电商行业遭受了严重的经济损失。准确的产品退货预测是供应商提前优化业务运营，降低退货相关成本的重要手段。本文提出了一种新的方法，称为contrasformer（对比变压器），通过利用从三种模式（即视觉、听觉和语言）中提取的细粒度流媒体行为特征，来预测实时流媒体电子商务中的产品回报。主要贡献在于我们采用了具有编码器-解码器架构的Transformer和一种新颖的类监督对比学习（CSCL），以融合多模态表示对齐和多模态交互表征的流媒体行为。通过使用从Tiktok中国直播平台收集的2584个产品流媒体和864个项目的真实数据集，我们证明了所提出的Contrasformer方法在预测产品退货率方面优于基线方法，平均绝对误差降低了25%。这项研究为供应商管理他们在直播商业中的实践提供了重要的管理启示。

{"title":"Product return prediction in live streaming e-commerce with cross-modal contrastive transformer","authors":"Wen Zhang , Rui Xie , Pei Quan , Zhenzhong Ma","doi":"10.1016/j.dss.2025.114470","DOIUrl":"10.1016/j.dss.2025.114470","url":null,"abstract":"<div><div>The live-streaming e-commerce industry is suffering heavy economic losses due to the high product return rate, which leads to rising logistics costs, greater inventory pressure, and unsatisfactory consumer experiences. Accurate product return prediction is highly desirable for the vendors to optimize their business operations in advance to reduce return-related costs. This paper proposes a novel approach, called Contraformer (Contrastive transformer), to predict product returns in live streaming e-commerce by leveraging fine-grained streamer behavior features extracted from three modalities (i.e., visual, acoustic, and language). The primary contribution lies in that we adopt Transformer with the encoder-decoder architecture with a novel class-supervised contrastive learning (CSCL) to fuse streamer behavior for multimodal representation alignment and inter-modal interaction characterization. By using a real-world dataset with 2584 product streamers and 864 items collected from Tiktok China live streaming platform, we demonstrate that the proposed Contrasformer approach outperforms the baseline methods in predicting product return rate with a 25 % reduction in terms of mean absolute error. This study offers great managerial implications for vendors to manage their practice in live streaming commerce.</div></div>","PeriodicalId":55181,"journal":{"name":"Decision Support Systems","volume":"194 ","pages":"Article 114470"},"PeriodicalIF":6.7,"publicationDate":"2025-05-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143922198","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0