Data & Knowledge Engineering最新文献

Privacy-preserving cross-network service recommendation via federated learning of unified user representations

IF 2.7 3区计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Data & Knowledge Engineering

Pub Date : 2025-03-04 DOI: 10.1016/j.datak.2025.102422

Mohamed Gaith Ayadi , Haithem Mezni , Hela Elmannai , Reem Ibrahim Alkanhel

With the emergence of cloud computing, the Internet of Things, and other large-scale environments, recommender systems have been faced with several issues, mainly (i) the distribution of user–item data across multiple information networks, (ii) privacy restrictions and the partial profiling of users and items caused by this distribution, (iii) the heterogeneity of user–item knowledge in different information networks. Furthermore, most approaches perform recommendations based on a single source of information, and do not handle the partial representation of users’ and items’ information in a federated way. Such isolated and non-collaborative behavior, in multi-source and cross-network information settings, often results in inaccurate and low-quality recommendations. To address these issues, we exploit the strengths of network representation learning and federated learning to propose a service recommendation approach in smart service networks. While NRL is employed to learn rich representations of entities (e.g., users, services, IoT objects), federated learning helps collaboratively infer a unified profile of users and items, based on the concept of anchor user, which are bridge entities connecting multiple information networks. These unified profiles are, finally, fed into a federated recommendation algorithm to select the top-rated services. Using a scenario from the smart healthcare context, the proposed approach was developed and validated on a multiplex information network built from real-world electronic medical records (157 diseases, 491 symptoms, 273 174 patients, treatments and anchors data). Experimental results under varied federated settings demonstrated the utility of cross-client knowledge (i.e. anchor links) and the collaborative reconstruction of composite embeddings (i.e. user representations) for improving recommendation accuracy. In terms of RMSE@K and MAE@K, our approach achieved an improvement of 54.41% compared to traditional single-network recommendation, as long as the federation and communication scale increased. Moreover, the gap with four federated approaches has reached 19.83 %, highlighting our approach’s ability to map local embeddings (i.e. user’s partial representations) into a complete view.

{"title":"Privacy-preserving cross-network service recommendation via federated learning of unified user representations","authors":"Mohamed Gaith Ayadi , Haithem Mezni , Hela Elmannai , Reem Ibrahim Alkanhel","doi":"10.1016/j.datak.2025.102422","DOIUrl":"10.1016/j.datak.2025.102422","url":null,"abstract":"<div><div>With the emergence of cloud computing, the Internet of Things, and other large-scale environments, recommender systems have been faced with several issues, mainly (i) the distribution of user–item data across multiple information networks, (ii) privacy restrictions and the partial profiling of users and items caused by this distribution, (iii) the heterogeneity of user–item knowledge in different information networks. Furthermore, most approaches perform recommendations based on a single source of information, and do not handle the partial representation of users’ and items’ information in a federated way. Such isolated and non-collaborative behavior, in multi-source and cross-network information settings, often results in inaccurate and low-quality recommendations. To address these issues, we exploit the strengths of network representation learning and federated learning to propose a service recommendation approach in smart service networks. While NRL is employed to learn rich representations of entities (e.g., users, services, IoT objects), federated learning helps collaboratively infer a unified profile of users and items, based on the concept of <em>anchor user</em>, which are bridge entities connecting multiple information networks. These unified profiles are, finally, fed into a federated recommendation algorithm to select the top-rated services. Using a scenario from the smart healthcare context, the proposed approach was developed and validated on a multiplex information network built from real-world electronic medical records (157 diseases, 491 symptoms, 273 174 patients, treatments and anchors data). Experimental results under varied federated settings demonstrated the utility of cross-client knowledge (i.e. anchor links) and the collaborative reconstruction of composite embeddings (i.e. user representations) for improving recommendation accuracy. In terms of RMSE@K and MAE@K, our approach achieved an improvement of 54.41% compared to traditional single-network recommendation, as long as the federation and communication scale increased. Moreover, the gap with four federated approaches has reached 19.83 %, highlighting our approach’s ability to map local embeddings (i.e. user’s partial representations) into a complete view.</div></div>","PeriodicalId":55184,"journal":{"name":"Data & Knowledge Engineering","volume":"158 ","pages":"Article 102422"},"PeriodicalIF":2.7,"publicationDate":"2025-03-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143551137","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

Advances in knowledge discovery and management, best papers of EGC 2024

IF 2.7 3区计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Data & Knowledge Engineering

Pub Date : 2025-03-01 DOI: 10.1016/j.datak.2025.102420

Jérôme Gensel , Christophe Cruz , Hocine Cherifi

引用次数: 0

CriMOnto: A generalized domain-specific ontology for modeling procedural norms of the Lebanese criminal law

IF 2.7 3区计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Data & Knowledge Engineering

Pub Date : 2025-02-27 DOI: 10.1016/j.datak.2025.102419

Mirna El Ghosh , Hala Naja , Habib Abdulrab , Mohamad Khalil

Criminal (or penal) law regulates offenses, offenders, and legal punishments. Modeling criminal law is gaining much attention in the ontology engineering community. However, a significant aspect is neglected: the explicit representation of procedural knowledge. Procedural norms, such as regulative norms, are addressed to agents in the normative system. They govern the different interactions among these agents. In this study, we propose a formal and faithful representation of the procedural aspect of legal norms in the context of the Lebanese Criminal Code. A modular domain-specific ontology named CriMOnto is developed for this purpose. CriMOnto is grounded in the Unified Foundational Ontology (UFO) and the legal core ontology UFO-L by applying the Ontology-Driven Conceptual Modeling (ODCM) process. Conceptual Ontology Patterns (COPs) are reused from UFO and UFO-L to build the hierarchical and procedural content of the ontology. CriMOnto is validated as a formal ontology and evaluated using a dual evaluation approach. The potential use of CriMOnto for lightweight rule-based decision support is discussed in this study.

{"title":"CriMOnto: A generalized domain-specific ontology for modeling procedural norms of the Lebanese criminal law","authors":"Mirna El Ghosh , Hala Naja , Habib Abdulrab , Mohamad Khalil","doi":"10.1016/j.datak.2025.102419","DOIUrl":"10.1016/j.datak.2025.102419","url":null,"abstract":"<div><div>Criminal (or penal) law regulates offenses, offenders, and legal punishments. Modeling criminal law is gaining much attention in the ontology engineering community. However, a significant aspect is neglected: the explicit representation of procedural knowledge. Procedural norms, such as regulative norms, are addressed to agents in the normative system. They govern the different interactions among these agents. In this study, we propose a formal and faithful representation of the procedural aspect of legal norms in the context of the Lebanese Criminal Code. A modular domain-specific ontology named CriMOnto is developed for this purpose. CriMOnto is grounded in the Unified Foundational Ontology (UFO) and the legal core ontology UFO-L by applying the Ontology-Driven Conceptual Modeling (ODCM) process. Conceptual Ontology Patterns (COPs) are reused from UFO and UFO-L to build the hierarchical and procedural content of the ontology. CriMOnto is validated as a formal ontology and evaluated using a dual evaluation approach. The potential use of CriMOnto for lightweight rule-based decision support is discussed in this study.</div></div>","PeriodicalId":55184,"journal":{"name":"Data & Knowledge Engineering","volume":"158 ","pages":"Article 102419"},"PeriodicalIF":2.7,"publicationDate":"2025-02-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143529128","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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 MDA approach for robotic-based real-time business intelligence applications

IF 2.7 3区计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Data & Knowledge Engineering

Pub Date : 2025-02-21 DOI: 10.1016/j.datak.2025.102418

Houssam Bazza , Sandro Bimonte , Zakaria Gourti , Stefano Rizzi , Hassan Badir

Industry 4.0, the fourth industrial revolution, has emerged from the convergence of robotics, automation, and the Internet of Things (IoT), transforming industrial processes with intelligent systems and digital integration. This revolution also brings with it Business Intelligence (BI) systems that enable the analysis of IoT and robotic data. The data architectures employed for BI in Industry 4.0 contexts are often intricate, typically comprising robots software, DBMSs, message brokers, and data stream management systems. Consequently, designing BI data-centric applications for Industry 4.0 presents a significant challenge. Inspired by the absence of modeling approaches for this type of application and by the well-established advantages of Model-Driven Architecture (MDA), this paper introduces a novel UML profile for real-time robotic data-driven BI applications. Our profile enables the representation of robotic and transactional data within a unified and consistent framework, enabling continuous queries over these streams. Additionally, we propose an automated method to implement UML class diagrams onto a technological stack featuring ROS, Apache Kafka, PostgreSQL, and Apache Flink. An experimental evaluation in the agricultural application domain confirms the merits of our approach.

{"title":"An MDA approach for robotic-based real-time business intelligence applications","authors":"Houssam Bazza , Sandro Bimonte , Zakaria Gourti , Stefano Rizzi , Hassan Badir","doi":"10.1016/j.datak.2025.102418","DOIUrl":"10.1016/j.datak.2025.102418","url":null,"abstract":"<div><div>Industry 4.0, the fourth industrial revolution, has emerged from the convergence of robotics, automation, and the Internet of Things (IoT), transforming industrial processes with intelligent systems and digital integration. This revolution also brings with it Business Intelligence (BI) systems that enable the analysis of IoT and robotic data. The data architectures employed for BI in Industry 4.0 contexts are often intricate, typically comprising robots software, DBMSs, message brokers, and data stream management systems. Consequently, designing BI data-centric applications for Industry 4.0 presents a significant challenge. Inspired by the absence of modeling approaches for this type of application and by the well-established advantages of Model-Driven Architecture (MDA), this paper introduces a novel UML profile for real-time robotic data-driven BI applications. Our profile enables the representation of robotic and transactional data within a unified and consistent framework, enabling continuous queries over these streams. Additionally, we propose an automated method to implement UML class diagrams onto a technological stack featuring ROS, Apache Kafka, PostgreSQL, and Apache Flink. An experimental evaluation in the agricultural application domain confirms the merits of our approach.</div></div>","PeriodicalId":55184,"journal":{"name":"Data & Knowledge Engineering","volume":"157 ","pages":"Article 102418"},"PeriodicalIF":2.7,"publicationDate":"2025-02-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143471503","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

Aspect-level recommendation fused with review and rating representations

IF 2.7 3区计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Data & Knowledge Engineering

Pub Date : 2025-02-18 DOI: 10.1016/j.datak.2025.102417

Heng-Ru Zhang , Ling Lin , Fan Min

Review contains user opinions about different aspects of an item, which is essential data for aspect-level recommendation. Most existing aspect-level recommendation algorithms are concerned with the degree to which user and item aspects match. However, even if an item is extremely popular due to its high quality, it may only partially match the aspects of a user. A tolerant user may like the item, whereas a strict user may dislike it. This implies that these works disregard the personalized behavior patterns of the user. In this paper, we propose a new Aspect-level Recommendation model fused with Review and Rating, namely ARRR, to address the recommendation bias. First, we introduce rating to explore user behavior patterns and item quality. Then, we present a personalized attention mechanism that generates a set of aspect-level user or item representations from reviews. Finally, we obtain comprehensive user or item representations by combining rating- and review-based representations. In the experiments, the proposed model is compared with seven state-of-the-art recommendation algorithms on seven datasets. The results show that our model outperforms on seven metrics. The source code of ARRR is available at https://github.com/alinn00/ARRR.

评论包含用户对物品不同方面的意见，是方面级推荐的重要数据。大多数现有的方面级推荐算法关注的是用户和项目方面的匹配程度。然而，即使一个项目因其高质量而非常受欢迎，它也可能只与用户的部分方面相匹配。宽容的用户可能喜欢这个项目，而严格的用户可能不喜欢它。这意味着这些作品忽略了用户的个性化行为模式。在本文中，我们提出了一种融合了评论和评级的新的方面级推荐模型，即 ARRR，以解决推荐偏差问题。首先，我们引入评分来探索用户行为模式和项目质量。然后，我们提出了一种个性化关注机制，该机制可从评论中生成一组方面级用户或项目表征。最后，我们通过结合评分和基于评论的表征来获得全面的用户或项目表征。在实验中，我们将所提出的模型与七种数据集上最先进的推荐算法进行了比较。结果表明，我们的模型在七个指标上都优于其他推荐算法。ARRR 的源代码见 https://github.com/alinn00/ARRR。

{"title":"Aspect-level recommendation fused with review and rating representations","authors":"Heng-Ru Zhang , Ling Lin , Fan Min","doi":"10.1016/j.datak.2025.102417","DOIUrl":"10.1016/j.datak.2025.102417","url":null,"abstract":"<div><div>Review contains user opinions about different aspects of an item, which is essential data for aspect-level recommendation. Most existing aspect-level recommendation algorithms are concerned with the degree to which user and item aspects match. However, even if an item is extremely popular due to its high quality, it may only partially match the aspects of a user. A tolerant user may like the item, whereas a strict user may dislike it. This implies that these works disregard the personalized behavior patterns of the user. In this paper, we propose a new <strong>A</strong>spect-level <strong>R</strong>ecommendation model fused with <strong>R</strong>eview and <strong>R</strong>ating, namely <strong>ARRR</strong>, to address the recommendation bias. First, we introduce rating to explore user behavior patterns and item quality. Then, we present a personalized attention mechanism that generates a set of aspect-level user or item representations from reviews. Finally, we obtain comprehensive user or item representations by combining rating- and review-based representations. In the experiments, the proposed model is compared with seven state-of-the-art recommendation algorithms on seven datasets. The results show that our model outperforms on seven metrics. The source code of ARRR is available at <span><span>https://github.com/alinn00/ARRR</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":55184,"journal":{"name":"Data & Knowledge Engineering","volume":"157 ","pages":"Article 102417"},"PeriodicalIF":2.7,"publicationDate":"2025-02-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143445354","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

Evaluating diabetes dataset for knowledge graph embedding based link prediction

IF 2.7 3区计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Data & Knowledge Engineering

Pub Date : 2025-02-15 DOI: 10.1016/j.datak.2025.102414

Sushmita Singh, Manvi Siwach

For doing any accurate analysis or prediction on data, a complete and well-populated dataset is required. Medical based data for any disease like diabetes is highly coupled and heterogeneous in nature, with numerous interconnections. This inherently complex data cannot be analysed by simple relational databases making knowledge graphs an ideal tool for its representation which can efficiently handle intricate relationships. Thus, knowledge graphs can be leveraged to analyse diabetes data, enhancing both the accuracy and efficiency of data-driven decision-making processes. Although substantial data exists on diabetes in various formats, the availability of organized and complete datasets is limited, highlighting the critical need for creation of a well-populated knowledge graph. Moreover while developing the knowledge graph, an inevitable problem of incompleteness is present due to missing links or relationships, necessitating the use of knowledge graph completion tasks to fill in this absent information which involves predicting missing data with various Link Prediction (LP) techniques. Among various link prediction methods, approaches based on knowledge graph embeddings have demonstrated superior performance and effectiveness. These knowledge graphs can support in-depth analysis and enhance the prediction of diabetes-associated risks in this field. This paper introduces a dataset specifically designed for performing link prediction on a diabetes knowledge graph, so that it can be used to fill the information gaps further contributing in the domain of risk analysis in diabetes. The accuracy of the dataset is assessed through validation with state-of-the-art embedding-based link prediction methods.

{"title":"Evaluating diabetes dataset for knowledge graph embedding based link prediction","authors":"Sushmita Singh, Manvi Siwach","doi":"10.1016/j.datak.2025.102414","DOIUrl":"10.1016/j.datak.2025.102414","url":null,"abstract":"<div><div>For doing any accurate analysis or prediction on data, a complete and well-populated dataset is required. Medical based data for any disease like diabetes is highly coupled and heterogeneous in nature, with numerous interconnections. This inherently complex data cannot be analysed by simple relational databases making knowledge graphs an ideal tool for its representation which can efficiently handle intricate relationships. Thus, knowledge graphs can be leveraged to analyse diabetes data, enhancing both the accuracy and efficiency of data-driven decision-making processes. Although substantial data exists on diabetes in various formats, the availability of organized and complete datasets is limited, highlighting the critical need for creation of a well-populated knowledge graph. Moreover while developing the knowledge graph, an inevitable problem of incompleteness is present due to missing links or relationships, necessitating the use of knowledge graph completion tasks to fill in this absent information which involves predicting missing data with various Link Prediction (LP) techniques. Among various link prediction methods, approaches based on knowledge graph embeddings have demonstrated superior performance and effectiveness. These knowledge graphs can support in-depth analysis and enhance the prediction of diabetes-associated risks in this field. This paper introduces a dataset specifically designed for performing link prediction on a diabetes knowledge graph, so that it can be used to fill the information gaps further contributing in the domain of risk analysis in diabetes. The accuracy of the dataset is assessed through validation with state-of-the-art embedding-based link prediction methods.</div></div>","PeriodicalId":55184,"journal":{"name":"Data & Knowledge Engineering","volume":"157 ","pages":"Article 102414"},"PeriodicalIF":2.7,"publicationDate":"2025-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143419097","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

How well can a large language model explain business processes as perceived by users?

IF 2.7 3区计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Data & Knowledge Engineering

Pub Date : 2025-02-15 DOI: 10.1016/j.datak.2025.102416

Dirk Fahland , Fabiana Fournier , Lior Limonad , Inna Skarbovsky , Ava J.E. Swevels

Large Language Models (LLMs) are trained on a vast amount of text to interpret and generate human-like textual content. They are becoming a vital vehicle in realizing the vision of the autonomous enterprise, with organizations today actively adopting LLMs to automate many aspects of their operations. LLMs are likely to play a prominent role in future AI-augmented business process management systems (ABPMSs) catering functionalities across all system lifecycle stages. One such system’s functionality is Situation-Aware eXplainability (SAX), which relates to generating causally sound and yet human-interpretable explanations that take into account the process context in which the explained condition occurred.

In this paper, we present the SAX4BPM framework developed to generate SAX explanations. The SAX4BPM suite consists of a set of services and a central knowledge repository. The functionality of these services is to elicit the various knowledge ingredients that underlie SAX explanations. A key innovative component among these ingredients is the causal process execution view. In this work, we integrate the framework with an LLM to leverage its power to synthesize the various input ingredients for the sake of improved SAX explanations.

Since the use of LLMs for SAX is also accompanied by a certain degree of doubt related to its capacity to adequately fulfill SAX along with its tendency for hallucination and lack of inherent capacity to reason, we pursued a methodological evaluation of the perceived quality of the generated explanations. To this aim, we developed a designated scale and conducted a rigorous user study. Our findings show that the input presented to the LLMs aided with the guard-railing of its performance, yielding SAX explanations having better-perceived fidelity. This improvement is moderated by the perception of trust and curiosity. More so, this improvement comes at the cost of the perceived interpretability of the explanation.

大型语言模型（LLM）是在大量文本基础上进行训练，以解释和生成类似人类的文本内容。大型语言模型正在成为实现自主企业愿景的重要工具，如今各组织都在积极采用大型语言模型来自动处理业务的许多方面。在未来的人工智能增强型业务流程管理系统（ABPMS）中，LLM 很可能会发挥重要作用，满足所有系统生命周期阶段的功能需求。情境感知可解释性（Situation-Aware eXplainability，SAX）是此类系统的功能之一，它与生成因果合理且可由人类解释的解释有关，其中考虑到了解释条件发生时的流程上下文。SAX4BPM 套件由一组服务和一个中央知识库组成。这些服务的功能是获取作为 SAX 解释基础的各种知识要素。这些要素中的一个关键创新组件是因果流程执行视图。在这项工作中，我们将该框架与 LLM 集成在一起，利用 LLM 综合各种输入成分的能力来改进 SAX 解释。由于在 SAX 中使用 LLM 也伴随着一定程度的质疑，即 LLM 是否有能力充分实现 SAX，以及 LLM 是否具有幻觉倾向和缺乏内在推理能力，因此我们对生成的解释的感知质量进行了方法学评估。为此，我们制定了一个指定的量表，并进行了严格的用户研究。我们的研究结果表明，向 LLMs 提供的输入有助于提高其性能，从而使 SAX 解释具有更好的保真度。这种改进受到信任感和好奇心的影响。更重要的是，这种改善是以解释的可解释性为代价的。

{"title":"How well can a large language model explain business processes as perceived by users?","authors":"Dirk Fahland , Fabiana Fournier , Lior Limonad , Inna Skarbovsky , Ava J.E. Swevels","doi":"10.1016/j.datak.2025.102416","DOIUrl":"10.1016/j.datak.2025.102416","url":null,"abstract":"<div><div>Large Language Models (LLMs) are trained on a vast amount of text to interpret and generate human-like textual content. They are becoming a vital vehicle in realizing the vision of the autonomous enterprise, with organizations today actively adopting LLMs to automate many aspects of their operations. LLMs are likely to play a prominent role in future AI-augmented business process management systems (ABPMSs) catering functionalities across all system lifecycle stages. One such system’s functionality is Situation-Aware eXplainability (SAX), which relates to generating causally sound and yet human-interpretable explanations that take into account the process context in which the explained condition occurred.</div><div>In this paper, we present the SAX4BPM framework developed to generate SAX explanations. The SAX4BPM suite consists of a set of services and a central knowledge repository. The functionality of these services is to elicit the various knowledge ingredients that underlie SAX explanations. A key innovative component among these ingredients is the causal process execution view. In this work, we integrate the framework with an LLM to leverage its power to synthesize the various input ingredients for the sake of improved SAX explanations.</div><div>Since the use of LLMs for SAX is also accompanied by a certain degree of doubt related to its capacity to adequately fulfill SAX along with its tendency for hallucination and lack of inherent capacity to reason, we pursued a methodological evaluation of the perceived quality of the generated explanations. To this aim, we developed a designated scale and conducted a rigorous user study. Our findings show that the input presented to the LLMs aided with the guard-railing of its performance, yielding SAX explanations having better-perceived fidelity. This improvement is moderated by the perception of trust and curiosity. More so, this improvement comes at the cost of the perceived interpretability of the explanation.</div></div>","PeriodicalId":55184,"journal":{"name":"Data & Knowledge Engineering","volume":"157 ","pages":"Article 102416"},"PeriodicalIF":2.7,"publicationDate":"2025-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143445243","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Modelling and solving industrial production tasks as planning-scheduling tasks

IF 2.7 3区计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Data & Knowledge Engineering

Pub Date : 2025-02-14 DOI: 10.1016/j.datak.2025.102415

Andrii Nyporko , Lukáš Chrpa

Industrial production planning or manufacturing concerns the selection of activities that can produce a desired product and scheduling them on resources that perform these activities. To deal with such problems techniques in the fields of Automated Planning and Scheduling might be leveraged, which are usually pursued separately even though they are (very) complementary. In manufacturing, the activities represent elementary steps in the production and each activity requires a specific input in order to produce a desired output. From that perspective, activities resemble actions in planning as they can capture such a requirement. Selecting proper activities including their (partial) ordering can be understood as a planning task while allocating the activities to the resources can be understood as a scheduling task.

This paper formalises the concept of “combined” planning and scheduling tasks by defining planning-scheduling tasks that are suitable for problems concerning industrial production or manufacturing. In particular, we define two types of activities – production and maintenance activities – where the former describes elementary production tasks while the latter modifies attributes of the resources (e.g. changing the configuration of reconfigurable machines). We introduce an extension of Planning Domain Definition Language (PDDL), a well-known language for describing planning tasks, to support modelling of planning-scheduling tasks. To tackle planning-scheduling tasks we propose two compilation schemes, one into temporal planning (in PDDL 2.1) and one into classical planning. We evaluated our approaches in three use cases of industrial production planning — Reconfigurable Machines, Woodworking, and Tube Factory domains. The results showed that solving planning-scheduling tasks by compiling them into planning tasks in order to use off-the-shelf planning engines is suitable as it scales reasonably well with the size of the actual tasks (although the resulting solutions are suboptimal).

{"title":"Modelling and solving industrial production tasks as planning-scheduling tasks","authors":"Andrii Nyporko , Lukáš Chrpa","doi":"10.1016/j.datak.2025.102415","DOIUrl":"10.1016/j.datak.2025.102415","url":null,"abstract":"<div><div>Industrial production planning or manufacturing concerns the selection of activities that can produce a desired product and scheduling them on resources that perform these activities. To deal with such problems techniques in the fields of <em>Automated Planning</em> and <em>Scheduling</em> might be leveraged, which are usually pursued separately even though they are (very) complementary. In manufacturing, the activities represent elementary steps in the production and each activity requires a specific input in order to produce a desired output. From that perspective, activities resemble actions in planning as they can capture such a requirement. Selecting proper activities including their (partial) ordering can be understood as a planning task while allocating the activities to the resources can be understood as a scheduling task.</div><div>This paper formalises the concept of “combined” planning and scheduling tasks by defining <em>planning-scheduling tasks</em> that are suitable for problems concerning industrial production or manufacturing. In particular, we define two types of activities – <em>production</em> and <em>maintenance</em> activities – where the former describes elementary production tasks while the latter modifies attributes of the resources (e.g. changing the configuration of reconfigurable machines). We introduce an extension of Planning Domain Definition Language (PDDL), a well-known language for describing planning tasks, to support modelling of planning-scheduling tasks. To tackle planning-scheduling tasks we propose two compilation schemes, one into temporal planning (in PDDL 2.1) and one into classical planning. We evaluated our approaches in three use cases of industrial production planning — Reconfigurable Machines, Woodworking, and Tube Factory domains. The results showed that solving planning-scheduling tasks by compiling them into planning tasks in order to use off-the-shelf planning engines is suitable as it scales reasonably well with the size of the actual tasks (although the resulting solutions are suboptimal).</div></div>","PeriodicalId":55184,"journal":{"name":"Data & Knowledge Engineering","volume":"157 ","pages":"Article 102415"},"PeriodicalIF":2.7,"publicationDate":"2025-02-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143437007","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

Static and dynamic techniques for iterative test-driven modelling of Dynamic Condition Response Graphs

IF 2.7 3区计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Data & Knowledge Engineering

Pub Date : 2025-02-04 DOI: 10.1016/j.datak.2025.102413

Axel K.F. Christfort , Vlad Paul Cosma , Søren Debois , Thomas T. Hildebrandt , Tijs Slaats

Test-driven declarative process modelling combines process models with test traces and has been introduced as a means to achieve both the flexibility provided by the declarative approach and the comprehensibility of the imperative approach. Open test-driven modelling adds a notion of context to tests, specifying the activities of concern in the model, and has been introduced as a means to support both iterative test-driven modelling, where the model can be extended without having to change all tests, and unit testing, where tests can define desired properties of parts of the process without needing to reason about the details of the whole process. The openness however makes checking a test more demanding, since actions outside the context are allowed at any point in the test execution and therefore many different traces may validate or invalidate an open test. In this paper we combine previously developed static techniques for effective open test-driven modelling for Dynamic Condition Response Graphs with a novel efficient implementation of dynamic checking of open tests based on alignment checking. We illustrate the static techniques on an example based on a real-life cross-organizational case management system and benchmark the dynamic checking on models and tests of varying size.

{"title":"Static and dynamic techniques for iterative test-driven modelling of Dynamic Condition Response Graphs","authors":"Axel K.F. Christfort , Vlad Paul Cosma , Søren Debois , Thomas T. Hildebrandt , Tijs Slaats","doi":"10.1016/j.datak.2025.102413","DOIUrl":"10.1016/j.datak.2025.102413","url":null,"abstract":"<div><div>Test-driven declarative process modelling combines process models with test traces and has been introduced as a means to achieve both the flexibility provided by the declarative approach and the comprehensibility of the imperative approach. Open test-driven modelling adds a notion of context to tests, specifying the activities of concern in the model, and has been introduced as a means to support both iterative test-driven modelling, where the model can be extended without having to change all tests, and unit testing, where tests can define desired properties of parts of the process without needing to reason about the details of the whole process. The openness however makes checking a test more demanding, since actions outside the context are allowed at any point in the test execution and therefore many different traces may validate or invalidate an open test. In this paper we combine previously developed static techniques for effective open test-driven modelling for Dynamic Condition Response Graphs with a novel efficient implementation of dynamic checking of open tests based on alignment checking. We illustrate the static techniques on an example based on a real-life cross-organizational case management system and benchmark the dynamic checking on models and tests of varying size.</div></div>","PeriodicalId":55184,"journal":{"name":"Data & Knowledge Engineering","volume":"157 ","pages":"Article 102413"},"PeriodicalIF":2.7,"publicationDate":"2025-02-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143377553","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"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

Reinforcement learning for optimizing responses in care processes

IF 2.7 3区计算机科学 Q3 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Data & Knowledge Engineering

Pub Date : 2025-02-03 DOI: 10.1016/j.datak.2025.102412

Olusanmi A. Hundogan , Bart J. Verhoef , Patrick Theeven , Hajo A. Reijers , Xixi Lu

Prescriptive process monitoring aims to derive recommendations for optimizing complex processes. While previous studies have successfully used reinforcement learning techniques to derive actionable policies in business processes, care processes present unique challenges due to their dynamic and multifaceted nature. For example, at any stage of a care process, a multitude of actions is possible. In this study, we follow the Reinforcement Learning (RL) approach and present a general approach that uses event data to build and train Markov decision processes. We proposed three algorithms including one that takes the elapsed time into account when transforming an event log into a semi-Markov decision process. We evaluated the RL approach using an aggression incident data set. Specifically, the goal is to optimize staff member actions when clients are displaying different types of aggressive behavior. The Q-learning and SARSA are used to find optimal policies. Our results showed that the derived policies align closely with current practices while offering alternative options in specific situations. By employing RL in the context of care processes, we contribute to the ongoing efforts to enhance decision-making and efficiency in dynamic and complex environments.

{"title":"Reinforcement learning for optimizing responses in care processes","authors":"Olusanmi A. Hundogan , Bart J. Verhoef , Patrick Theeven , Hajo A. Reijers , Xixi Lu","doi":"10.1016/j.datak.2025.102412","DOIUrl":"10.1016/j.datak.2025.102412","url":null,"abstract":"<div><div>Prescriptive process monitoring aims to derive recommendations for optimizing complex processes. While previous studies have successfully used reinforcement learning techniques to derive actionable policies in business processes, care processes present unique challenges due to their dynamic and multifaceted nature. For example, at any stage of a care process, a multitude of actions is possible. In this study, we follow the Reinforcement Learning (RL) approach and present a general approach that uses event data to build and train Markov decision processes. We proposed three algorithms including one that takes the elapsed time into account when transforming an event log into a semi-Markov decision process. We evaluated the RL approach using an aggression incident data set. Specifically, the goal is to optimize staff member actions when clients are displaying different types of aggressive behavior. The Q-learning and SARSA are used to find optimal policies. Our results showed that the derived policies align closely with current practices while offering alternative options in specific situations. By employing RL in the context of care processes, we contribute to the ongoing efforts to enhance decision-making and efficiency in dynamic and complex environments.</div></div>","PeriodicalId":55184,"journal":{"name":"Data & Knowledge Engineering","volume":"157 ","pages":"Article 102412"},"PeriodicalIF":2.7,"publicationDate":"2025-02-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143372710","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0