Skip to main content

2024 | Book

Cooperative Information Systems

29th International Conference, CoopIS 2023, Groningen, The Netherlands, October 30–November 3, 2023, Proceedings

Editors: Mohamed Sellami, Maria-Esther Vidal, Boudewijn van Dongen, Walid Gaaloul, Hervé Panetto

Publisher: Springer Nature Switzerland

Book Series : Lecture Notes in Computer Science


About this book

This book constitutes the refereed proceedings of the 29th International Conference on Cooperative Information Systems, CoopIS 2023, held in Groningen, The Netherlands, during October 30–November 3, 2023. The 21 regular papers and 10 work-in-progress papers included in this book were carefully reviewed and selected from 100 submissions. They were organized in topical sections as follows: Knowledge Engineering; Deployment and Migration in CISs; Security and Privacy in CISs; Process Modeling; Process Analytics; Human Aspects and Social Interaction in CISs; and Work in Progress.

Table of Contents


Knowledge Engineering

Enhancing Fairness and Accuracy in Machine Learning Through Similarity Networks

Machine Learning is a powerful tool for uncovering relationships and patterns within datasets. However, applying it to a large datasets can lead to biased outcomes and quality issues, due to confounder variables indirectly related to the outcome of interest. Achieving fairness often alters training data, like balancing imbalanced groups (privileged/unprivileged) or excluding sensitive features, impacting accuracy. To address this, we propose a solution inspired by similarity network fusion, preserving dataset structure by integrating global and local similarities. We evaluate our method, considering data set complexity, fairness, and accuracy. Experimental results show the similarity network’s effectiveness in balancing fairness and accuracy. We discuss implications and future directions.

Samira Maghool, Elena Casiraghi, Paolo Ceravolo
Considering Vocabulary Mappings in Query Plans for Federations of RDF Data Sources

Federations of RDF data sources offer great potential for queries that cannot be answered by a single data source. However, querying such federations poses several challenges, one of which is that different but semantically-overlapping vocabularies may be used for the respective RDF data. Since the federation members usually retain their autonomy, this heterogeneity cannot simply be homogenized by modifying the data in the data sources. Therefore, handling this heterogeneity becomes a critical aspect of query planning and execution. We introduce an approach to address this challenge by leveraging vocabulary mappings for the processing of queries over federations with heterogeneous vocabularies. This approach not only translates SPARQL queries but also preserves the correctness of results during query execution. We demonstrate the effectiveness of the approach and measure how the application of vocabulary mappings affects on the performance of federated query processing.

Sijin Cheng, Sebastián Ferrada, Olaf Hartig
AIS - A Metric for Assessing the Impact of an Influencer’s Twitter Activity on the Price of a Cryptocurrency

Individual users on social media platforms like Twitter can significantly volatile assets, including cryptocurrencies. However, current research has overlooked this aspect, focusing on sentiment analysis that includes all posts from all users. Making it challenging to detect trends caused by individuals. To address this gap, we introduce the Asset Influence Score (AIS), a percentage-based metric that assesses the likelihood of a newly issued tweet aligning with periods of heightened trading activity. By analyzing price data and tweets concurrently, we identify correlations that enable to predict the likelihood of specific users’ tweets co-occurring with increased trading activity. Evaluating the AIS using a publicly available prototype and Twitter data from 2020 to 2023, we find that using the AIS as a buy signal outperforms buy-and-hold and technical trading strategies while maintaining high liquidity. Demonstrating the applicability of AIS in improving trading decisions and identifying key individuals on social media platforms.

Kevin Miller, Kristof Böhmer

Deployment and Migration in CISs

Managing the Variability of Component Implementations and Their Deployment Configurations Across Heterogeneous Deployment Technologies

Application systems often need to be deployed in different variants if requirements that influence their implementation, hosting, and configuration differ between customers. Therefore, deployment technologies, such as Ansible or Terraform, support a certain degree of variability modeling. Besides, modern application systems typically consist of various software components deployed using multiple deployment technologies that only support their proprietary, non-interoperable variability modeling concepts. The Variable Deployment Metamodel (VDMM) manages the deployment variability across heterogeneous deployment technologies based on a single variable deployment model. However, VDMM currently only supports modeling conditional components and their relations which is sometimes too coarse-grained since it requires modeling entire components, including their implementation and deployment configuration for each different component variant. Therefore, we extend VDMM by a more fine-grained approach for managing the variability of component implementations and their deployment configurations, e.g., if a cheap version of a SaaS deployment provides only a community edition of the software and not the enterprise edition, which has additional analytical reporting functionalities built-in. We show that our extended VDMM can be used to realize variable deployments across different individual deployment technologies using a case study and our prototype OpenTOSCA Vintner.

Miles Stötzner, Uwe Breitenbücher, Robin D. Pesl, Steffen Becker
Adaptive Multi-agent System for Dynamic Clustering Applied to Itineraries Regularities and Traffic Prediction

Nowadays, electronic devices such as mobile phones, sensors embedded in vehicles, and more generally digital acquisition devices continuously provide information relating the state of the surrounding environment at nearly real-time. These data provide important information required for real-time systems monitoring such as temperature regulation in smart building, traffic planning to relieve network congestion in smart cities, etc. The growing demand for analysing such data encourages researchers to adopt an approach known as “streaming analysis” which aims at processing data streams at real-time to continuously capture data evolution over time. Data stream clustering approaches already exist but require keeping in memory all the input data and have a slow adaptation to data changes. To solve these problems, we propose to agentify the clusters and allow them to fuse and evolve locally and autonomously. This work presents AMAS4DC a generic Dynamic Clustering model based on Adaptive Multi-Agent System approach. AMAS4DC processes acquired data on the fly using local similarity evaluation for cluster’s creation or fusion. AMAS4DC is then instantiated on two use cases: the dynamic clustering of itineraries to detect regularities and the dynamic clustering of traffic data to predict future traffic. The conducted experiments underline the performance of AMAS4DC in terms of memory usage, processing time and clustering quality compared to well-known models for dynamic clustering.

Alexandre Perles, Ha Nhi Ngo, Elsy Kaddoum, Valérie Camps
Double Deep Q-Network-Based Time and Energy-Efficient Mobility-Aware Workflow Migration Approach

With the emergence of the Fog paradigm, the relocation of computational capabilities to the network’s edge has become imperative to support the ever-growing requirements of latency-sensitive, data-intensive, and real-time decision-making applications. In dynamic and mobile environments, services must adapt to accommodate the mobility of users, resulting in frequent relocations across computing nodes to ensure seamless user experiences. However, these migrations incur additional costs and potentially degrade the Quality of Service (QoS) parameters. In this paper, we propose a mobility-aware workflow migration approach based on Deep Reinforcement Learning (DRL). This approach aims to minimize the system’s overall delay and energy consumption by optimizing the number of workflow task migrations, considering resource performance and network conditions in different regions. The problem is first formulated as a Markov Decision Process (MDP), and then a Double Deep Q-network (DDQN) algorithm is proposed to identify the optimal policy for workflow offloading and migration. Comprehensive experiments have been conducted and the results demonstrate that our approach outperforms significantly the existing approaches.

Nour El Houda Boubaker, Karim Zarour, Nawal Guermouche, Djamel Benmerzoug

Security and Privacy in CISs

Decentralized and Autonomous Key Management for Open Multi-agent Systems of Embedded Agents

This paper presents a public key infrastructure for open multi-agent systems of embedded agents. Open multi-agent systems of embedded agents are a set of network embedded systems cooperating in real time to achieve their goal without a central server issuing commands. In this context, agents are very prone to attacks as they can be confronted with new agents with unknown goals and as they often rely on wireless ad hoc communications. The key infrastructure we propose allows the agents to communicate without the risk of their messages being tampered with. Thus, providing foundations for more advanced security solutions such as trust management systems. In order to do that, we leverage the agent’s capabilities to establish and maintain the infrastructure by providing self-organization and trust management rules. Once established, the infrastructure provides a way for the agents to assert their right to take part in the system operations with certificates and to secure their communications with asymmetric cryptography. As a result, agents can communicate securely without the risk of their identities being stolen and their communications being tampered with. They are also being able to exclude intruders. The work proposed in this paper paves the way to build more secure open decentralized systems of autonomous embedded systems. To make our solution general and adaptable to many situations, we decoupled the cryptographic and trust management details from the infrastructure itself.

Arthur Baudet, Oum-El-Kheir Aktouf, Annabelle Mercier, Philippe Elbaz-Vincent
An Empirical Study on Socio-technical Modeling for Interdisciplinary Privacy Requirements

Data protection regulations impose requirements on organizations that require interdisciplinary. Conceptual modeling of information systems, particularly goal modeling, has served to communicate with stakeholders of different backgrounds for software requirements analysis. An extension for a Socio-Technical Security (STS) modeling language was proposed to include data protection modeling concepts to help represent relevant issues of the European Union’s General Data Protection Regulation. This article examines whether models designed with this extension serve as communication facilitators for privacy compliance and common ground across stakeholders.Through a series of 8 focus groups, with 21 subjects, we observed if professionals with different backgrounds (software developers, business analysts, and privacy experts) could detect discuss about the GDPR principles and identify privacy compliance “red flags” that we seeded in a use case. Using a qualitative approach to analyze the data, all the groups discussed the majority of the GDPR principles and identified more than 80% of the seeded red flags, with privacy experts identifying the most. This research provides preliminary results on using conceptual modeling as a communicator facilitator between stakeholders to contribute to a common ground between them.

Claudia Negri-Ribalta, Rene Noel, Oscar Pastor, Camille Salinesi
Enhancing Workflow Security in Multi-cloud Environments Through Monitoring and Adaptation upon Cloud Service and Network Security Violations

Cloud computing has emerged as a crucial solution for handling data- and compute-intensive workflows, offering scalability to address dynamic demands. However, ensuring the secure execution of workflows in the untrusted multi-cloud environment poses significant challenges, given the sensitive nature of the involved data and tasks. The lack of comprehensive approaches for detecting attacks during workflow execution, coupled with inadequate measures for reacting to security and privacy breaches has been identified in the literature. To close this gap, in this work, we propose an approach that focuses on monitoring cloud services and networks to detect security violations during workflow executions. Upon detection, our approach selects the optimal adaptation action to minimize the impact on the workflow. To mitigate the uncertain cost associated with such adaptations and their potential impact on other tasks in the workflow, we employ adaptive learning to determine the most suitable adaptation action. Our approach is evaluated based on the performance of the detection procedure and the impact of the selected adaptations on the workflows.

Nafiseh Soveizi, Dimka Karastoyanova

Process Modeling

Beyond Rule-Based Named Entity Recognition and Relation Extraction for Process Model Generation from Natural Language Text

Process-aware information systems offer extensive advantages to companies, facilitating planning, operations, and optimization of day-to-day business activities. However, the time-consuming but required step of designing formal business process models often hampers the potential of these systems. To overcome this challenge, automated generation of business process models from natural language text has emerged as a promising approach to expedite this step. Generally two crucial subtasks have to be solved: extracting process-relevant information from natural language and creating the actual model. Approaches towards the first subtask are rule based methods, highly optimized for specific domains, but hard to adapt to related applications. To solve this issue, we present an extension to an existing pipeline, to make it entirely data driven. We demonstrate the competitiveness of our improved pipeline, which not only eliminates the substantial overhead associated with feature engineering and rule definition, but also enables adaptation to different datasets, entity and relation types, and new domains. Additionally, the largest available dataset (PET) for the first subtask, contains no information about linguistic references between mentions of entities in the process description. Yet, the resolution of these mentions into a single visual element is essential for high quality process models. We propose an extension to the PET dataset that incorporates information about linguistic references and a corresponding method for resolving them. Finally, we provide a detailed analysis of the inherent challenges in the dataset at hand.

Julian Neuberger, Lars Ackermann, Stefan Jablonski
LABPMN: Location-Aware Business Process Modeling and Notation

The combination of IoT and BPM enables new possibilities for the use of contextual information during the modeling and execution of process models. Nevertheless, many approaches for the use of location data only exist as concepts, and most existing extensions for BPMN do not fully use the potential gained through IoT.In this paper, we introduce a novel BPMN Extension for location awareness that is, conceptually well-defined and adheres to the BPMN meta-model, and can be used both graphically, and during the execution of a process. We introduce two new main elements: The possibility of dynamically assigning or allocating actors to one or a number of tasks, and different location-based events to be able to react to location changes of active and passive resources during process execution.

Leo Poss, Lukas Dietz, Stefan Schönig
On the Semantic Transparency of Declarative Process Models: The Case of Constraints

Process modeling notations are essential for analyzing, designing, improving, and digitalizing business processes in organizations. In particular, for knowledge-intensive processes, notations that allow flexible orchestration of activities are crucial for supporting discretionary work. Declarative Process Modeling notations allow describing the interplay between the process activities with constraints that need to be met, without specifying how to meet them. Despite being well-suited for modeling flexible processes, declarative notations are not as widely adopted as imperative notations where the process is usually depicted using a flow-based approach. This paper focuses on investigating the semantic transparency of declarative notations, specifically how the visual representation of constraints aligns with the underlying formal concepts used in Declarative Process Models (DPMs). The study concentrates on Dynamic Condition Response (DCR) Graphs, a representative notation of DPMs extensively used in industry and academia. The research employs semi-structured interviews with experts in DCR Graphs, as well as an analysis of semantic transparency based on theoretical models of understandability. The findings indicate that generally colors contribute to the understanding of relations in DPMs, while the shapes used to describe constraints do not accurately convey their semantics. Based on these results, the study proposes an alternative representation of constraints, paving the road for an enhanced representation of DPMs.

Dung My Thi Trinh, Amine Abbad-Andaloussi, Hugo A. López

Process Analytics

Discovering Guard Stage Milestone Models Through Hierarchical Clustering

Processes executed on enterprise Information Systems (IS), such as ERP and CMS, are artifact-centric. The execution of these processes is driven by the creation and evolution of business entities called artifacts. Several artifact-centric modeling languages were proposed to capture the specificity of these processes. One of the most used artifact-centric modeling languages is the Guard Stage Milestone (GSM) language. It represents an artifact-centric process as an information model and a lifecycle. The lifecycle groups activities in stages with data conditions as guards. The hierarchy between the stages is based on common conditions. However, existing works do not discover this hierarchy nor the data conditions, as they considered them to be already available. They also do not discover GSM models directly from event logs. They discover Petri nets and translate them into GSM models. To fill this gap, we propose in this paper a discovery approach based on hierarchical clustering. We use invariants detection to discover data conditions and information gain of common conditions to cluster stages. The approach does not rely on domain knowledge nor translation mechanisms. It was implemented and evaluated using a blockchain case study.

Leyla Moctar M’Baba, Mohamed Sellami, Nour Assy, Walid Gaaloul, Mohamedade Farouk Nanne
Discovery of Workflow Patterns - A Comparison of Process Discovery Algorithms

Process mining provides a set of techniques and algorithms to analyze, support, and improve business processes based on process execution data. Process discovery aims at deducing a representative process model of real-world execution. So far, process discovery algorithms have been mainly compared regarding their output quality but not yet with regard to their functional capabilities. The well-established workflow control flow patterns imperatively describe process behavior, originally used to compare modeling languages, but to date, not to compare discovery algorithms. In this work, we analyze a representative set of process discovery algorithms with regard to their coverage of 23 control flow patterns. For this purpose, we implemented each workflow pattern as an executable colored Petri net, simulated it, and ran various discovery algorithms on the obtained event log. A comparison of the results shows that the discovery algorithms mainly cover basic control flow patterns and iterative structures, while multi-instance, state-base, and cancellation patterns are only partially covered.

Kerstin Andree, Mai Hoang, Felix Dannenberg, Ingo Weber, Luise Pufahl
From Process Mining Insights to Process Improvement: All Talk and No Action?

Organizations from various domains use process mining to better understand, analyze, and improve their business processes. While the overall value of process mining has been shown in several contexts, little is known about the specific actions that are taken to move from process mining insights to process improvement. In this work, we address this research gap by conducting a systematic literature review. Specifically, we investigate which types of actions have been taken in existing studies and to which insights these actions are linked. Our findings show that there exists a large variety of actions. Many of these actions do not only relate to changes to the investigated process but also to the associated information systems, the process documentation, the communication between staff members, and personnel training. Understanding the diversity of the actions triggered by process mining insights is important to instigate future research on the different aspects of translating process mining insights into process improvement. The insights-to-action realm presented in this work can inform and inspire new process mining initiatives and prepare for the effort required after acquiring process mining insights.

Vinicius Stein Dani, Henrik Leopold, Jan Martijn E. M. van der Werf, Iris Beerepoot, Hajo A. Reijers
Rectify Sensor Data in IoT: A Case Study on Enabling Process Mining for Logistic Process in an Air Cargo Terminal

The Internet of Things (IoT) has empowered enterprises to optimize process efficiency and productivity by analyzing sensor data. This can be achieved with process mining, a technology that enables organizations to extract valuable insights from data recorded during process execution, referred to as event data in a process mining context. In our case study, we aim to apply process mining to sensor data collected within a logistic process at an air cargo terminal, specifically from device-to-device communication. By representing the sensor data as event data, we rectify them to accurately capture the movement of package distribution in the logistic process. However, due to the communication dynamics, challenges arise from the presence of irrelevant data that does not impact the process instance’s status. Moreover, issues such as faulty sensor readings and ambiguous data interpretation further compound these challenges. To overcome the obstacles, we collaborate with domain experts to develop rules that take into account the context of each event in a trace, enabling us to effectively capture package distribution within the system. We present the results of our process mining analysis, which have been validated by domain experts. This case study contributes to the understanding and utilization of sensor data for process mining in IoT environments, with a specific focus on data collected from device-to-device communication.

Chiao-Yun Li, Aparna Joshi, Nicholas T. L. Tam, Sean Shing Fung Lau, Jinhui Huang, Tejaswini Shinde, Wil M. P. van der Aalst
Using Process Mining for Face Validity Assessment in Agent-Based Simulation Models: An Exploratory Case Study

In the field of simulation, the key objective of a system designer is to develop a model that performs a specific task and accurately represents real-world systems or processes. A valid simulation model allows for a better understanding of the system’s behavior and improved decision-making in the real world. Face validity is a subjective measure that assesses the extent to which a simulation model and its outcomes appear reasonable to an expert based on a superficial examination of the simulator’s realism. Process mining techniques, which are novel data-driven methods for obtaining real-life insights into processes based on event logs, show promise when combined with effective visualization techniques. These techniques can augment the face validity assessment of simulation models in reflecting real-life behavior and play a key role in supporting humans conducting such assessments. In this paper, we present an approach that utilizes process mining techniques to assess the face validity of agent-based simulation models. To illustrate our approach, we use the Schelling model of segregation. We demonstrate how graphical representation, immersive assessment, and sensitivity analysis can be used to assess face validity based on event logs produced by the simulation model. Our study shows that process mining in combination with visualization can strongly support humans in assessing face validity of agent-based simulation models.

Rob Bemthuis, Ruben Govers, Sanja Lazarova-Molnar

Human Aspects and Social Interaction in CISs

Towards Scaling External Feedback for Early-Stage Researchers: A Survey Study

Feedback on research artefacts from people beyond local research groups, such as researchers in online research communities, has the potential to bring in additional support for early-stage researchers and complementary viewpoints to research projects. While current literature has focused primarily on early-stage research seeking or getting support for research skills development in general, less is known about, more specifically, empirical understanding of how early-stage researchers exchange feedback with external researchers. In this paper, we focus on understanding the critical types of external feedback that early-stage researchers desire and the prevalent challenges they face with exchanging feedback with external helpers. We report on a large-scale survey conducted with early-stage researchers of diverse backgrounds. Our findings lay the empirical foundation for informing the designing of socio-technical systems for research feedback exchange.

Yuchao Jiang, Marcos Báez, Boualem Benatallah
Social Network Mining from Natural Language Text and Event Logs for Compliance Deviation Detection

Social network mining aims at discovering and visualizing information exchange of resources and relations of resources among each other. For this, most existing approaches consider event logs as input data and therefore only depict how work was performed (as-is) and neglect information on how work should be performed (to-be), i.e., whether or not the actual execution is in compliance with the execution specified by the company or law. To bridge this gap, the presented approach considers event logs and natural language texts as input outlining rules on how resources are supposed to work together and which information may be exchanged between them. For pre-processing the natural language texts the large language model GPT-4 is utilized and its output is fed into a customized organizational mining component which delivers the to-be organizational perspective. In addition, we integrate well-known process discovery techniques from event logs to gather the as-is perspective. A comparison in the form of a graphical representation of both, the to-be and as-is perspectives, enables users to detect deviating behavior. The approach is evaluated based on a set of well-established process descriptions as well as synthetic and real-world event logs.

Henryk Mustroph, Karolin Winter, Stefanie Rinderle-Ma
Learning Hierarchical Robot Skills Represented by Behavior Trees from Natural Language

Learning from natural language is a programming-free and user friendly teaching method that allows users without programming knowledge or demonstration capabilities to instruct robots, which has great value in industry and daily life. The manipulation skills of robots are often hierarchical skills composed of low-level primitive skills, so they can be conveniently represented by behavior trees (BTs). Based on this idea, we propose NL2BT, a framework for generating behavior trees from natural language and controlling robots to complete hierarchical tasks in real time. The framework consists of two language processing stages, an initial behavior tree library composed of primitive skill subtrees, and a BT-Generation algorithm. To validate the effectiveness of NL2BT, we use it to build a Chinese natural language system for instructing robots in performing 3C assembly tasks, which is a significant application of Industry 4.0. We also discuss the positive impact of real-time teaching, visual student models, and the synonymous skill module in the framework. In addition to the demonstrated application, NL2BT can be easily migrated to other languages and hierarchical task learning scenarios.

Kaiyi Wang, Yongjia Zhao, Shuling Dai, Minghao Yang, Yichen He, Ning Zhang
Relating Context and Self Awareness in the Internet of Things

Context- and self- awareness are two terms that have been living with us for many years. In principle, both state a similar meaning even though the literature points out a very different path. One is inspired by location-related mechanisms in mobile environments, whereas the other is inspired by biology. In the area of the Internet of Things, the term context-awareness has seen a higher adoption in the field of Cloud Computing, while the term self-awareness is more widely used in the area of Wireless Sensor Networks. This paper evaluates the entire IoT Cloud-to-Thing Continuum in an attempt to reconcile both terms. We contextualize and discuss the literature around context and self-awareness, and we propose a conceptual architecture that handles both concepts, with the aim of having a better understanding of how to develop a software environment that integrates both concepts. To show the real-life applicability of our proposed architecture, it is introduced in a realistic setting such as wildfire monitoring, including a conceptual overview of how the proposed architecture could be implemented in this domain. Additionally, our evaluation of a river flooding scenario concluded that the proposed architecture significantly reduced flood detection delay by over 47% compared to the naive method and over 20% compared to standalone self-awareness and context-awareness methods.

David Arnaiz, Marc Vila, Eduard Alarcón, Francesc Moll, Maria-Ribera Sancho, Ernest Teniente

Work in Progress (WIP) Papers

BAnDIT: Business Process Anomaly Detection in Transactions

Business process anomaly detection enables the prevention of misuse and failures. Existing approaches focus on detecting anomalies in control, temporal, and resource behavior of individual instances, neglecting the communication of multiple instances in choreographies. Consequently, anomaly detection capabilities are limited. This study presents a novel neural network-based approach to detect anomalies in distributed business processes. Unlike existing methods, our solution considers message data exchanged during process transactions. Allowing the generation of detection profiles incorporating the relationship between multiple instances, related services, and exchanged data to detect point and contextual anomalies during process runtime. To validate the proposed solution, it is demonstrated with a prototype implementation and validated with a use case from the ecommerce domain. Future work aims to further improve the deep learning approach, to enhance detection performance.

Nico Rudolf, Kristof Böhmer, Maria Leitner
Resource-Driven Process Manipulation: Modeling Concepts and Valid Allocations

In situations of scarce resource availability, flexibility on which resources execute which tasks is key to process and system performance. Tightly coupled control flow and resource modeling hampers flexible resource allocation. Hence, in this work, we propose resource-driven process manipulation (RDPM) to enable the separation between the business and resource requirements for a process. RDPM enables process modelers to specify resource-specific requirements for the control flow as part of resource profiles, e.g., a machine (resource) requires configuration (task) before execution. Moreover, the resource is promoted to a first-class citizen in process-aware information systems and enabled to impact the execution. The basic concepts of RDPM are defined and an algorithm is provided to find valid resource allocations for a task. The approach is prototypically implemented and compared to existing modeling approaches w.r.t. complexity for the modeler and process participant.

Felix Schumann, Stefanie Rinderle-Ma
Graph Collaborative Filtering and Data Augmentation Strategies in Dual-Target CDR

Current dual-target recommendation methods focus on efficient feature fusion but neglect the inherent noise issues in the domains. However, noise negatively affects the fusion of domains. To tackle this issue, we introduce an improvement and noise reduction strategy named DA-DCDR(Data Augmentation-Dual Target Cross Domain Recommendation), for domain fusion. By refining and reducing noise in each domain’s subgraphs, we not only enhance the accuracy of interaction data but also ensure consistency in data scales. To establish associations between distinct domains, we implement a graph co-training strategy. Key procedures of DA-DCDR include interaction refinement and noise reduction, domain fusion, and correlation expansion. We use graph encoders to acquire user/item embeddings for both domains before domain fusion, followed by enhancement and noise reduction in interactions via top-k sampling and re-prediction. Additionally, we amplify user-user and item-item correlation elements after the domain fusion. Experimental results validate the noteworthy performance enhancement of our proposed strategy in the dual-target recommendation, mitigating the noise effects and boosting the accuracy of the dual-target recommendation system.

Xiaowen Shao, Baisong Liu, Xueyuan Zhang, Junru Li, Ercong Xu, Shiqi Wu
Clustering Raw Sensor Data in Process Logs to Detect Data Streams

The execution and analysis of processes is strongly influenced by sensor streams, e.g., temperature, that are measured in parallel to the process execution and stored in process event logs. This holds particularly true for application domains such as logistics and manufacturing. However, currently, these sensor streams are collected and stored in an arbitrary and unsystematic way. Hence, this work proposes an approach that prepares sensor streams into individual data streams that can be annotated to process tasks and used for process analysis and prediction.

Matthias Ehrendorfer, Juergen Mangler, Stefanie Rinderle-Ma
Comparing the Performance of GPT-3 with BERT for Decision Requirements Modeling

Operational decisions such as loan or subsidy allocation are taken with high frequency and require a consistent decision quality which decision models can ensure. Decision models can be derived from textual descriptions describing both the decision logic and decision dependencies. Whilst decision models already help with modeling, implementing and automating decisions, the modelling step would still benefit from a (semi)-automated approach. The introduction of ChatGPT and GPT-3 offers opportunities to automatically discover decision dependencies from a given text. This paper evaluates the performance of two approaches that automatically extract decision dependencies from text, namely the best performing version of GPT-3 with a BERT-based approach. An evaluation with 36 experiments with a dataset of real-life cases and various levels of creativity allowed for GPT-3 concludes that theBERT BERT-based approach outperforms GPT-3 on the real-life dataset but that GPT-3 has promising results and requires further investigation.

Alexandre Goossens, Johannes De Smedt, Jan Vanthienen
A Requirements Study on Model Repositories for Digital Twins in Construction Engineering

Building information modeling is becoming the preferred tool-assisted methodology in civil and construction engineering for the design, management, and creation of digital replicas of buildings. However, current tool support for creating and managing these twins is limited as of lacking interoperability at the model level. Yet, the Industry foundation classes describe a standardized format for exchanging models and data of buildings and bears a strong resemblance to the class- and object-based nature of the Unified modeling language. From this resemblance, we postulate the application of model-driven engineering for establishing a model repository as an open collaboration platform in Building information modeling. Based on our experience from ongoing and concluded interdisciplinary research projects in civil and construction engineering and computer science, in this paper, we seminally elicit requirements for such a model repository.

Philipp Zech, Georg Fröch, Ruth Breu
Joint Dynamic Resource Allocation and Trajectory Optimization for UAV-Assisted Mobile Edge Computing in Internet of Vehicles

Computation offloading in Mobile Edge Computing (MEC) represents a key technology for the future of the Internet of Vehicles (IoV), reducing the time and energy consumption of vehicles for computation tasks, while Unmanned Aerial Vehicles (UAVs) equipped with computation resources can act as aerial based stations to provide computation offloading services to vehicles moving on the road. In this paper, a joint dynamic resource allocation and UAV trajectory optimization scheme is proposed. In the scheme, an UAV is deployed with an edge server to execute the partially offloaded computation tasks from multiple vehicles. The goal of the problem is to maximize the total computation workload while minimizing the energy consumption of all vehicles by jointly optimizing the computation frequency, the wireless transmission power, the task offloading decisions of vehicles, as well as the flight angle of the UAV in each time slot. Since the problem is non-convex in continuous action space, we consider the Twin Delayed Deep Deterministic (TD3) policy gradient algorithm to solve the problem. Experimental results demonstrate the effectiveness of the TD3 policy gradient algorithm in the proposed optimization scheme in terms of the convergence speed and the system reward.

Runji Li, Haifeng Sun
Towards an Improved Unsupervised Graph-Based MRI Brain Segmentation Method

Brain disorders are becoming more prevalent, and accurate brain segmentation is a vital component of identifying the appropriate treatment. This study introduces an enhanced graph-based image segmentation technique. The node selection process involves creating an ellipsoid centered at the image’s center of mass. The proposed approach is evaluated using the NFBS dataset and demonstrates superior visual and numerical outcomes compared to some of existing approaches.

Maria Popa, Anca Andreica
User-Friendly Exploration of Highly Heterogeneous Data Lakes

The proliferation of digital data sources and formats has led to the apparition of data lakes, systems where numerous data sources coexist, with less (or no) control and coordination among the sources, than previously practised in enterprise databases and data warehouses. While most data lakes are designed for very large number of tables, ConnectionLens [2, 3] is a data lake system for structured, semi-structured, and unstructured data, which it integrates into a single graph; the graph can be explored via graph queries with keyword search [4] and entity path enumeration [5]. In this paper, we describe ConnectionStudio, a user-friendly platform leveraging ConnectionLens, and integrating feedback from non-expert users, in particular, journalists. Our main insights are: (i) improve and entice exploration by giving a first global view; (ii) facilitate tabular exports from the integrated graph; (iii) provide interactive means to improve the graph constructions. The insights can be used to further advance the exploration and usage of data lakes for non-IT users.

Nelly Barret, Simon Ebel, Théo Galizzi, Ioana Manolescu, Madhulika Mohanty
Optimizing Hospital Patient Flow by Predicting Aftercare Requests from Fuzzy Time Series

Predictive modelling can be a huge benefit when it comes to optimizing patient flows in a hospital. Hospital beds are considered critical resources, thus the need for optimizing patient flow is evident. This paper focuses on predicting the out-flow of hospital patients to external aftercare facilities, to mitigate the waiting times that currently dominate this flow and have a negative influence on the patient recovery process. In order to achieve this, we analyze hospital patient time series data in the form of aftercare requests. Such predictions allow hospital and aftercare facilities to be aligned such that, as soon as a patient is medically ready for discharge, the aftercare facility can immediately allocate the patient, avoiding for such patient to stay longer in the hospital occupying a bed while waiting for a place in the aftercare facility.

Renata M. de Carvalho, Stef van der Sommen, Danilo F. de Carvalho
Cooperative Information Systems
Mohamed Sellami
Maria-Esther Vidal
Boudewijn van Dongen
Walid Gaaloul
Hervé Panetto
Copyright Year
Electronic ISBN
Print ISBN

Premium Partner