Cooperative Information Systems

In high connectivity knowledge graph, distance based knowledge graph embedding methods show promising performance on link prediction task, and are capable of encoding complex relations and key relation patterns. However, the existing methods fail to achieve excellent results in knowledge graph with poor context structure information. To mitigate this problem, we propose Bi2E, a bidirectional model based on subject-object feature spaces. To enhance the efficiency of data utilization and perceive more potential semantic links, we utilize the bidirectionality of relation to model from both forward and reverse directions. And Bi2E represents triples in the subject-object feature spaces, which enables it to capture richer feature information from rare data. In addition, Bi2E employs adaptive margin \(\gamma \) which makes embedded representation more flexible by only using a small amount of feature information. Experiments on link prediction benchmarks demonstrate the proposed key capabilities of Bi2E. Moreover, we set a new state-of-the-art on two low connectivity knowledge graph benchmarks.

In the latest years, the progressive digitalisation of Smart City ecosystems has fuelled an increasing availability of data from sensor networks, which is considered as a valuable asset for improving mobility resilience. In particular, data coming from sensors in vehicles can be leveraged to obtain useful information about the quality of the area-wide road surface in near real-time, and may be used by road maintainers to focus monitoring and maintenance activities on urban and public infrastructure. To bring such application scenario into the field, road maintainers should be equipped with valuable tools to gain insights from the data and ensure a safer and more efficient infrastructure. In this paper, we present a methodological approach, based on big data exploration techniques, applied to support road maintainers in analysing and assessing surface conditions of roads. Specifically, the proposed approach is grounded on three components: (i) a multi-dimensional model, apt to represent the road network and to enable data exploration; (ii) data summarisation techniques, in order to simplify overall view over high volumes of data collected by vehicles; (iii) a measure of relevance, aimed at focusing the attention of the maintainers on relevant data only. The paper illustrates the design and implementation of multiple exploration scenarios on top of the three components and their implementation and preliminary evaluation in an ongoing research project on sustainable and resilient mobility.

Opacity is a property of information flow that characterizes the ability of a system to keep a secret information hidden from a malicious external entity, called an attacker. Given a critical system that may leak confidential information, an attacker with partial observation of the system and a subset of controllable actions, we propose an approach to synthesize a controller that enforces the system’s opacity. This controller is designed as a function that applies, at run time, to the current execution to disable any controllable action that eventually leads to the violation of the opacity of the system. The supervision function is built at design time based on a new version of the symbolic observation graph that represents a reduced abstraction of the state space graph of the system preserving the observation of both the attacker and the controller. The language induced by this function is proven to be controllable, observable and supremal no matter the relation that exists between the observations of the attacker and the controller.

In recent years, the volume of data to be analyzed has increased tremendously. However, purposeful data analyses on large-scale data require in-depth domain knowledge. A common approach to reduce data volume and preserve interactivity are sampling algorithms. However, when using a sample, the semantic context across the entire dataset is lost, which impedes data preprocessing. In particular data imputation transformations, which aim to fill empty values for more accurate data analyses, suffer from this problem. To cope with this issue, we introduce DATA-IMP, a novel human-in-the-loop approach that enables data imputation transformations in an interactive manner while preserving scalability. We implemented a fully working prototype and conducted a comprehensive user study as well as a comparison to several non-interactive data imputation techniques. We show that our approach significantly outperforms state-of-the-art approaches regarding accuracy as well as preserves user satisfaction and enables domain experts to preprocess large-scale data in an interactive manner.

Process mining employs event data extracted from different types of information systems to discover and analyze actual processes. Event data often contain highly sensitive information about the people who carry out activities or the people for whom activities are performed. Therefore, privacy concerns in process mining are receiving increasing attention. To alleviate privacy-related risks, several privacy preservation techniques have been proposed. Differential privacy is one of these techniques which provides strong privacy guarantees. However, the proposed techniques presume that event data are released in only one shot, whereas business processes are continuously executed. Hence, event data are published repeatedly, resulting in additional risks. In this paper, we demonstrate that continuously released event data are not independent, and the correlation among different releases can result in privacy degradation when the same differential privacy mechanism is applied to each release. We quantify such privacy degradation in the form of temporal privacy leakages. We apply continuous event data publishing scenarios to real-life event logs to demonstrate privacy leakages.

In this paper, we propose a new ensemble method, which is called Dynamic Forest, for learning from data streams with varying feature spaces. Unlike traditional online learning where the feature space is static, in varying feature spaces, new features may emerge while others may vanish. This leads to several problems for which state-of-the-art online random forest algorithms are not equipped. We benchmark our proposed method against the state-of-the-art method OLVF on data streams with varying feature spaces and against OLVF and \(OL_{SF}\) on trapezoidal data streams. These trapezoidal data streams can be considered as a sub-problem of varying feature spaces, where the only characteristic is that new features emerge over time. Our proposed approach dynamically learns and relearns decision stumps while applying a dynamic weighting strategy for the decision stumps. Furthermore, it employs a dynamic strategy for adding and removing weak learners. The proposed method is empirically evaluated by replicating the benchmark of the OLVF algorithm with nine UCI Machine Learning Repository datasets and one real-world dataset. In the experiments, we can show that Dynamic Forest proves to be a good addition to the current state-of-the-art for learning from data streams with varying feature spaces.

With the abundance of event data, the challenge of enabling process discovery in the large has attracted the community attention. Several works addressed the problem by performing process discovery directly on relational databases, instead of the traditional file based computations. Preliminary results show that moving (parts of) process discovery to the database engine outperforms file based computations. However, all existing works consider the traditional storage of event data which assumes that a clear and predefined process instance notion exists, and that events are correlated to one process instance. In this work, we go two steps further. First, we address the problem of process discovery on object-centric event data which allows several process instance notions to be flexibly defined. We refer to it as multi-process discovery Second, motivated by the intrinsic nature of process discovery that searches for relationships in event data, we address the question of how graph-based storage of object-centric event data improves the performance of multi-process discovery? We propose in-database process discovery operators based on labeled property graphs. We use Neo4j as a DBMS and Cypher as a query language. We compare different discovery strategies that involve graph and relational databases. Our results show that process discovery in graph databases outperform existing approaches.

Collaborative work environments have gained more attention in manufacturing in recent years, in particular with the development of collaborative robots (cobots), a special type of industrial robot that is built for the safe interaction between humans and robots. Recent advances have shown that there are several collaboration scenarios between humans and robots, such as, synchronized, cooperation, collaboration and coexistence. So far, most literature focuses only on the collaboration between one human and one robot. However, literature also predicts that there will be more collaboration scenarios with one or many humans collaborating with one or many robots. Furthermore, literature on collaboration scenarios often focuses only on a generic process perspective and does not detail tasks nor other aspects. In this paper, we aim to address these gaps by investigating collaboration scenarios for one to many and many to many relations between robots and humans in workflows. First, we formalize the collaboration pattern and its types (synchronized, cooperation, collaboration and coexistence). Our approach allows for the specification of time-based, spatial and functional constraints at task level in collaborative work environments. Second, we demonstrate our findings with a proof-of-concept implementation that consists of a workflow system, a cobot simulation and a communication and data platform. Third, we evaluate our model with altogether seven use cases (e.g., spot taping). The results show that the patterns can be applied for the specification of collaboration scenarios in modern, process-oriented work environments. For future work, we would like to investigate questions on process modeling and visualization of collaborative patterns.

Graphs are becoming increasingly larger, with datasets having millions of vertices and billions (or even trillions) of edges. As a result, the ability to fit the entire graph into the main memory of a single machine faces challenges in common hardware, even more so in edge/IoT-like devices (i.e., more energy efficient but also more resource constrained). Reading the graph from secondary storage may pose in itself significant overhead, negatively impacting query performance and storage requirements. It thus becomes relevant to explore techniques to optimize the storage of graphs, specially in memory, in a way that circumvents space limitations, while avoiding compromising the performance of processing.

We observe that current graph storage systems manage the graph representation by storing graphs in an uncompressed format, either: i) in a shared architecture which leads to a higher space overhead and the inability to represent the graph entirely in main memory, or ii) in a distributed architecture, where the graph dataset is partitioned over a cluster of machines with each one storing in main memory only a fragment (shard) of the (uncompressed) graph. We present PK-Graph, our proposal which extends a distributed graph processing system, highly used in academia and industry (Spark GraphX), in order to deploy the use of a compressed graph representation, with added support for dynamic updatable graphs (not currently supported in GraphX). Our experimental results show that PK-Graph can achieve up to 50% lower graph memory usage, while maintaining competitive performance in executing typical graph operations used in common applications.

In many industries, competitors are required to cooperate in order to conduct optimizations, e.g., to solve an assignment problem. For example, in air traffic flow management (ATFM), flight prioritization in case of temporarily reduced capacity of the air traffic network is an instance of the assignment problem. Participants, however, are typically reluctant to share sensitive information regarding their preferences for the optimization, which renders conventional approaches to optimization inadequate. This paper proposes a method for combining genetic algorithms with multi-party computation (MPC) as the basis for building a platform for optimizing the assignment of resources to different agents under the assumption of an honest-but-curious platform provider; the method is illustrated on the ATFM use case. In the proposed method a genetic algorithm iteratively generates a population of candidate solutions to the assignment problem while a Privacy Engine component evaluates the population in each iteration step. The participants’ private inputs are kept from competitors and not even the platform provider knows those inputs, receiving only encrypted input which is processed by MPC nodes in a way that preserves the secrecy of the inputs.

Object-aware processes enable the data-driven generation of forms based on the object behavior, which is pre-specified by the respective object lifecycle process. Each state of a lifecycle process comprises a number of object attributes that need to be set (e.g., via forms) before transitioning to the next state. When initially modeling a lifecycle process, the optimal ordering of the form fields is often unknown and only a guess of the lifecycle process modeler. As a consequence, certain form fields might be obsolete, missing, or ordered in a non-intuitive manner. Though this does not affect process executability, it decreases the usability of the automatically generated forms. Discovering respective problems, therefore, provides valuable insights into how object- and process-aware information systems can be evolved to improve their usability. This paper presents an approach for deriving improvements of object lifecycle processes by comparing the respective positions of the fields of the generated forms with the ones according to which the fields were actually filled by users during runtime. Our approach enables us to discover missing or obsolete form fields, and additionally considers the order of the fields within the generated forms. Finally, we can derive the modeling operations required to automatically restructure the internal logic of the lifecycle process states and, thus, to automatically evolve lifecycle processes and corresponding forms.

Analyzing event logs generated during the execution of digital processes, organizations can monitor the behavior of dysfunctional or unspecified processes. For achieving the most refined results, high-quality and up-to-date process models are required. However, the selection of the proper process discovery algorithm is often addressed by human experts that can relate quality criteria, event logs behavior, and discovery techniques. Exploiting a meta-learning approach, we created a procedure that identifies the optimal discovery technique based on a user-defined balance of quality metrics. Our experiments exploited 1091 event logs representing extensive possible business process behaviors. Given a set of available algorithms, we obtained an F-score of 0.76 for recommending the discovery algorithm that maximizes quality criteria. Moreover, our method supports a more in-depth investigation of the process discovery problem by mapping log behavior and discovery techniques.

Blockchain has been proposed as a trusted execution system, ensuring business process execution integrity and transparency. Smart contracts can manage the task or workflow execution and the allocation of tasks in a decentralized and reliable fashion. Nonetheless, blockchain transactions are public and accessible to their participants, and the issue of privacy is a well-known issue of blockchain systems for business process management. In the example of a service payment occurring after a sealed-bid auction, participants may not be willing to reveal the value of the accepted bid to other competitors. In this paper, we leverage smart contracts and a bank that manages per-collaboration payment tokens. The tokens are backed with fiat money with a conversion rate that is kept secret between payment partners and the bank. Hence, partners benefit from the interests of smart contracts such as autonomous programmable payment while preserving the confidentiality of the payment value. We implement this protocol in a real-world setting to demonstrate the approach’s feasibility, and we carry on quantitative experiments to confirm the validity of the protocol.

The International Data Spaces Association (IDSA) has promoted the idea of International Data Spaces as a place for companies to share data with trust and security enforced by software and organizational competence. There has been considerable progress in delivering corporate guidelines, technical specifications, and software components available for testing and deploying applications to support IDS-based ecosystems, such as the IDS data connectors classified by the Fraunhofer Institute. However, full implementation of IDS applications seems still complex and expensive for small and medium enterprises (SMEs). A possible strategy to deal with such an issue is to break the IDSA specification’s complexity into smaller pieces and build small IDS ecosystems formed by its core business roles (e.g., data owners, users, and broker service providers). In this context, this paper addresses the problem of designing an application to support the broker service provider’s role in operating in an IDS-based ecosystem. This research, therefore, follows a Design Science approach in a three-step process. First, it investigates problems of practical relevance elicited from the IDSA guidelines in combination with requirements provided by representatives of the Dutch Logistics sector. Second, it gives design to tackle the problem by combining Semantic Web, Linked Data, and Enterprise Architecture modeling artifacts. Last, it validates the architecture of the broker service provider’s application by demonstrating its technical feasibility, innovation, and software integration.

Structural network analysis retrieves the holistic patterns of interactions among network instances. Due to the unprecedented growth of data availability, it is time to take advantage of Machine Learning to integrate the outcome of the structural analysis with better predictions on the upcoming states of large networks. Concerning the existing challenges of adopting methods embracing multi-dimensional, multi-task, transparent representations within incremental procedures, in our recent study, we proposed the AVPRA algorithm. It works as an embedder of both the network structure and domain-specific features making the aforementioned challenges feasible to address. In this paper, we elaborate on the validation of AVPRA by adopting it in multiple downstream Machine Learning tasks on the Twitter network of the Italian Parliament. Comparing the outcome with state-of-the-art algorithms of graph embedding, the capability of AVPRA in retaining either network structure properties or domain-specific features of the nodes is promising. In addition, the method is incremental and transparent.

We propose a novel system for creating data plugs and sockets for plug-and-play database web services. We adopt a plug-and-play approach to couple an application to a database. In our approach a designer constructs a “plug,” which is a simple specification of the output produced by the service. If the plug can be “played” on the database “socket” then the web service is generated. Our plug-and-play approach has three advantages. First, a plug is portable. A plug can be played on any data source to generate a web service. Second, a plug is reliable. The database is checked to ensure that the service can be safely and correctly generated. Third, plug-and-play web services are easier to code for complex data since a service designer can write a simple plug, abstracting away the data’s real complexity. We describe a system for plug-and-play web services and experimentally evaluate the system.

Ensuring the quality of education and training has always been the focus of social concern. Traditionally, many chain education and training institutions have adopted a centralized management platform, which will lead to a lack of effective supervision among institutions. If the training records of trainees have problems such as data falsification/tampering, it is not easy to trace back, so that the security and credibility of the training data cannot be guaranteed. Moreover, due to the influence of geographical, economic and other factors, there is a lack of unified assessment standards in various regions, which makes it difficult to guarantee the training quality of trainees. In this context, this paper proposes an education and training management system (ETS) based on Hyperledger Fabric blockchain technology to provide a safe and reliable traceability solution for education and training management. In this system, we define some rules through smart contracts to implement different business logic. Using the characteristics of decentralization and high credibility of the blockchain, it solves the problems of uneven training quality and untraceable training records in the traditional education and training process. The system we built enables reliable sharing and privacy protection of training data. In addition, this paper provides an effective network configuration idea to obtain the best performance of the blockchain system. The performance of the proposed system is evaluated by experiments.

Conformance checking deals with collating modeled process behavior with observed process behavior recorded in event data. Alignments are a state-of-the-art technique to detect, localize, and quantify deviations in process executions, i.e., traces, compared to reference process models. Alignments, however, assume complete process executions covering the entire process from start to finish or prefixes of process executions. This paper defines infix/postfix alignments, proposes approaches to their computation, and evaluates them using real-life event data.

Business process modeling languages support enterprises in visualizing workflows in a graphical representation. Many studies provide recommendations about which modeling language to choose and how to represent models in terms of usability. However, there is no support in how to present detailed instructions regarding the execution of process tasks. We denote such instructions as task annotations which have to be considered during process execution to ensure process success. Integrating this information in an understandable way into process models is challenging and has not been sufficiently researched. This paper describes a novel study to address how task annotations can be presented in process models intuitively. In an experimental setup, we compare different representation formats for different task settings and evaluate them regarding the aspects effectiveness, mental efficiency and satisfaction. We found empirical support that image- and diagram-based representations are intuitively comprehensible across all task settings regardless of the user’s level of experience or education. Furthermore, we could statistically prove inferiority of textual task annotations.

Knowledge-intensive processes (KiPs) progress in a flexible way towards the achievement of process goals. Contextual factors like location and regulations affect how these goals are achieved in KiPs. Conventionally, a context is considered to be either static or dynamic. For some KiPs part of the context can be dynamic, meaning that the context can change during the execution of the KiP as a result of the decisions and interpretations of the knowledge-worker based on the information gained throughout the process. A holistic approach linking dynamic context, goals and processes is vital for modeling such KiPs. This paper presents a method, based on enterprise models, for integrated modeling of KiPs with contextual goals under dynamic contexts. With our method, we guide business analysts in modeling complex, flexible KiPs under dynamic contexts.