Adversarial attacks pose a significant threat to machine learning models by inducing incorrect predictions through imperceptible perturbations to input data. While these attacks are well studied in unstructured domains such as images, their behaviour on tabular data remains underexplored due to mixed feature types and complex inter-feature dependencies. This study introduces a comprehensive benchmark that evaluates adversarial attacks on tabular datasets with respect to both effectiveness and imperceptibility. We assess five white-box attack algorithms (FGSM, BIM, PGD, DeepFool, and C&W) across four representative models (LR, MLP, TabTransformer and FT-Transformer) using eleven datasets spanning finance, energy, and healthcare domains. The benchmark employs four quantitative imperceptibility metrics (proximity, sparsity, deviation, and sensitivity) to characterise perturbation realism. The analysis quantifies the trade-off between these two aspects and reveals consistent differences between attack types, with ℓ ∞ -based attacks achieving higher success but lower subtlety, and ℓ 2 -based attacks offering more realistic perturbations. The benchmark findings offer actionable insights for designing more imperceptible adversarial attacks, advancing the understanding of adversarial vulnerability in tabular machine learning.

Process discovery aims to derive a process model that accurately represents the observed behavior in an event log. As a state-of-the-art process discovery technique, Inductive Miner (IM) generates sound process models (i.e., free of deadlocks) while ensuring optimal replay fitness. However, IM may sometimes produce over-generalized process models with locally imprecise structures, often resulting in the creation of so-called flower structures. To address this limitation, this paper presents a novel technique that refines the process model generated by IM by optimizing its imprecise sub-processes. Specifically, the technique begins by identifying and extracting sub-logs corresponding to imprecise sub-processes in the initial IM-generated process model. Then, these imprecise sub-processes are iteratively optimized using a frequency-based filtering mechanism applied to the sub-logs. Once optimized, the imprecise sub-processes in the initial process model are replaced by the optimized ones, generating a set of candidates process models. Finally, the candidate with the best quality, in terms of fitness and precision, is selected as the final optimized process models. The proposed technique has been implemented as a plugin for the open-source process mining platform ProM. Through comparisons with state-of-the-art process discovery techniques using 10 publicly available real-life event logs, the experimental results demonstrate that the proposed method achieves an average absolute improvement of 0.173 in F-measure over its IMi variant, while also exhibiting competitive performance relative to other state-of-the-art approaches.

Real-world processes involve multiple object types with intricate interrelationships. Traditional event logs (in XES format), which record process execution centred around the case notion, are restricted to a single-object perspective, making it difficult to capture the behaviour of multiple objects and their interactions. To address this limitation, object-centric event logs (OCEL) have been introduced to capture both the objects involved in a process and their interactions with events. The object-centric event data (OCED) metamodel extends the OCEL format by further capturing dynamic object attributes and object-to-object relations. Recently OCEL 2.0 has been proposed based on OCED metamodel. Current research on generating OCEL logs requires specific input data sources, and resulting log data often fails to fully conform to OCEL 2.0. Moreover, the generated OCEL logs vary across different representational formats and their quality remains unevaluated. To address these challenges, a set of quality criteria for evaluating OCEL log representations is established. Guided by these criteria, Dirigo is proposed—a method for extracting event logs that not only conforms to OCEL 2.0 but also extends it by capturing the temporal aspect of dynamic object-to-object relations. Object-role Modelling (ORM), a conceptual data modelling technique, is employed to describe the artifact produced at each step of Dirigo . To validate the applicability of Dirigo , it is applied to a real-life use case. The quality of the log representation of the extracted event log is compared to those of existing OCEL logs using the established quality criteria. • Established quality criteria for evaluating OCEL log representations. • Proposed a method to extract event logs that conforms to the latest OCEL 2.0 schema. • Demonstrated the method’s applicability through validation using a real-life use case.

Explainable Artificial Intelligence (XAI) techniques are used to provide transparency to complex, opaque predictive models. However, these techniques are often designed for image and text data, and it is unclear how fit-for-purpose they are when applied to tabular data. As XAI techniques are rarely evaluated in the context of tabular data, the applicability of existing evaluation criteria and methods are also unclear and needs re-examination. For example, some works suggest that evaluation methods may unduly influence the evaluation results when using tabular data. This lack of clarity on evaluation procedures can lead to reduced transparency and ineffective use of XAI techniques in real world settings. In this study, we examine literature on XAI evaluation to derive guidelines on functionally-grounded assessment of local, post hoc XAI techniques. We identify 20 evaluation criteria and associated evaluation methods, and derive guidelines on when and how each criterion should be evaluated. We also identify key research gaps to be addressed by future work. Our study contributes to the body of knowledge on XAI evaluation through in-depth examination of functionally-grounded XAI evaluation protocols, and has laid the groundwork for future research on XAI evaluation.

Predictive Process Monitoring (PPM) aims to predict the future state of running process instances to enable timely interventions to mitigate potential risks. As one of the most fundamental tasks in PPM, process remaining time prediction focuses on preventing timeout occurrences. While various deep learning-based approaches have been developed for this purpose, they often rely on pre-established static prediction models and struggle to maintain accurate predictions when the process undergoes dynamic changes, such as an expanding sales channels. To tackle this challenge, this paper proposes an incremental process remaining time prediction framework by continuously updating the prediction model based on an incremental event log. Specifically, a feature selection strategy is first introduced to extract effective features from event logs. Leveraging effective features can significantly improve the prediction quality by capturing the changes in process information. Then, three incremental log-based updating mechanisms, including period-based updating, quantity-based updating, and concept-drift-based updating, along with a reconstruction strategy, are proposed to dynamically adjust the prediction model in response to business changes. Finally, LSTM, Transformer, and Auto-encoder models are adapted and integrated into the proposed framework. The approach has been implemented and publicly released. Experimental evaluation using nine real-life event logs demonstrate that the proposed framework and its three instantiations (i.e., LSTM-based, Transformer-based, and Auto-encoder-based ones) outperform state-of-the-art techniques in terms of prediction accuracy.

In recent years, various machine and deep learning architectures have been successfully introduced to the field of predictive process analytics. Nevertheless, the inherent opacity of these algorithms poses a significant challenge for human decision-makers, hindering their ability to understand the reasoning behind the predictions. This growing concern has sparked the introduction of counterfactual explanations, designed as human-understandable ‘what if’ scenarios, to provide clearer insights into the decision-making process behind undesirable predictions. The generation of counterfactual explanations, however, encounters specific challenges when dealing with the sequential nature of the (business) process cases typically used in predictive process analytics. Our paper tackles this challenge by introducing a data-driven approach, REVISED$^{+}$, to generate more feasible and plausible counterfactual explanations. First, we restrict the counterfactual algorithm to generate counterfactuals that lie within a high-density region of the process data, ensuring that the proposed counterfactuals are realistic and feasible within the observed process data distribution. Additionally, we ensure plausibility by learning sequential patterns between the activities in the process cases, utilising Declare language templates. Finally, we evaluate the properties that define the validity of counterfactuals.

Deep learning (DL) in orthopaedics has gained significant attention in recent years. Previous studies have shown that DL can be applied to a wide variety of orthopaedic tasks, including fracture detection, bone tumour diagnosis, implant recognition, and evaluation of osteoarthritis severity. The utilisation of DL is expected to increase, owing to its ability to present accurate diagnoses more efficiently than traditional methods in many scenarios. This reduces the time and cost of diagnosis for patients and orthopaedic surgeons. To our knowledge, no exclusive study has comprehensively reviewed all aspects of DL currently used in orthopaedic practice. This review addresses this knowledge gap using articles from Science Direct, Scopus, IEEE Xplore, and Web of Science between 2017 and 2023. The authors begin with the motivation for using DL in orthopaedics, including its ability to enhance diagnosis and treatment planning. The review then covers various applications of DL in orthopaedics, including fracture detection, detection of supraspinatus tears using MRI, osteoarthritis, prediction of types of arthroplasty implants, bone age assessment, and detection of joint-specific soft tissue disease. We also examine the challenges for implementing DL in orthopaedics, including the scarcity of data to train DL and the lack of interpretability, as well as possible solutions to these common pitfalls. Our work highlights the requirements to achieve trustworthiness in the outcomes generated by DL, including the need for accuracy, explainability, and fairness in the DL models. We pay particular attention to fusion techniques as one of the ways to increase trustworthiness, which have also been used to address the common multimodality in orthopaedics. Finally, we have reviewed the approval requirements set forth by the US Food and Drug Administration to enable the use of DL applications. As such, we aim to have this review function as a guide for researchers to develop a reliable DL application for orthopaedic tasks from scratch for use in the market.

Organizations can benefit from the use of practices, techniques, and tools from the area of business process management. Through the focus on processes, they create process models that require management, including support for versioning, refactoring and querying. Querying thus far has primarily focused on structural properties of models rather than on exploiting behavioral properties capturing aspects of model execution. While the latter is more challenging, it is also more effective, especially when models are used for auditing or process automation. The focus of this paper is to overcome the challenges associated with behavioral querying of process models in order to unlock its benefits. The first challenge concerns determining decidability of the building blocks of the query language, which are the possible behavioral relations between process tasks. The second challenge concerns achieving acceptable performance of query evaluation. The evaluation of a query may require expensive checks in all process models, of which there may be thousands. In light of these challenges, this paper proposes a special-purpose programming language, namely Process Query Language (PQL) for behavioral querying of process model collections. The language relies on a set of behavioral predicates between process tasks, whose usefulness has been empirically evaluated with a pool of process model stakeholders. This study resulted in a selection of the predicates to be implemented in PQL, whose decidability has also been formally proven. The computational performance of the language has been extensively evaluated through a set of experiments against two large process model collections.

Predictive Process Monitoringaims to predict the future information of ongoing process executions by leveraging machine and deep learning techniques. One of the tasks is known asremaining time prediction, which focuses on predicting the remaining time of ongoing cases. Accurate remaining time prediction can be valuable and important for improving business operations or taking timely interventions to prevent delays. For predicting the remaining time, existing work has used deep learning techniques to achieve high prediction accuracy. However, most of these techniques tend to learn very complex models that are difficult to explain. Systematic feature selection approaches may help improve both the prediction accuracy and the explainability of the model. In this paper, we introduce a feature-informed cascade prediction framework to predict the remaining time. Specifically, we first propose an approach that builds a tree of features by systematically estimating their effects on the remaining time prediction. Next, we use the tree to either automatically select an optimal combination of features or to guide users in this selection process. Each selected feature is correlated with its prediction results in our Feature-informed Cascade Prediction Model (FCPM) for explainability. The proposed approach has been implemented and is made publicly available. Using eight public real-life event logs, the proposed approach is compared to the state-of-the-art approaches in terms of prediction accuracy. In addition, it is demonstrated that our approach visualizes the impact of each input feature in the prediction of individual cases, producing explanations of the prediction results.

Radiologists are trained professionals who use medical images to obtain clinically relevant information. However, little is known about visual search patterns and strategies radiologists employ during medical image analysis. Thus, there is a current need for guidelines to specify optimal visual search routines commonly used by radiologists. Identifying these features could improve radiologist training and assist radiologists in their work. Our study found that during the moments in which radiologists view chest X-ray images in silence before verbalizing the analysis, they exhibit unique search patterns regardless of the type of disease depicted. Our findings suggest that radiologists’ search behaviors can be identified at this stage. However, when radiologists verbally interpret the X-rays, the gaze patterns appear noisy and arbitrary. Current deep-learning approaches train their systems using this noisy and arbitrary gaze data. This may explain why previous research still needs to show the superiority of deep-learning models that use eye tracking for disease classification. Our paper investigates these patterns and attempts to uncover the eye-gaze configurations during the different analysis phases.

Process mining focuses on the analysis of event log data to build various process analytical capabilities. Predictive process analytics has emerged as one of such key capabilities and it uses machine learning techniques to construct process prediction models. In recent years, deep neural networks have gained increasing interest in process prediction since they can handle multi-dimensional sequential inputs with minimal information loss. However, they are considered black-box models and existing studies in explaining deep neural network-based process predictions rely on only event-level features for explanation. In this paper, we propose a new approach for generating explanations for process outcome predictions at multiple levels. The approach is underpinned by three different prediction models: a transparent model for generating global explanations based on case-level features, an attention-based deep neural network for generating local explanations based on event-level features, and a novel eXplainable Dual-learning Deep network (XD2-net) for generating local explanations based on case-level features. Using three publicly available datasets, we have tested the applicability of the approach and further examined the multi-level explanations generated by the approach through an elaborate case study. Unlike others, the design of our approach promotes the idea of leveraging the complementary capabilities of different models and utilizing their strengths, rather than focusing on model performance competition. This will contribute towards generating more comprehensive explanations that meet the needs of different end users and purposes in the future.

Given the tremendous potential and influence of artificial intelligence (AI) and algorithmic decision‐making (DM), these systems have found wide‐ranging applications across diverse fields, including education, business, healthcare industries, government, and justice sectors. While AI and DM offer significant benefits, they also carry the risk of unfavourable outcomes for users and society. As a result, ensuring the safety, reliability, and trustworthiness of these systems becomes crucial. This article aims to provide a comprehensive review of the synergy between AI and DM, focussing on the importance of trustworthiness. The review addresses the following four key questions, guiding readers towards a deeper understanding of this topic: (i) why do we need trustworthy AI? (ii) what are the requirements for trustworthy AI? In line with this second question, the key requirements that establish the trustworthiness of these systems have been explained, including explainability, accountability, robustness, fairness, acceptance of AI, privacy, accuracy, reproducibility, and human agency, and oversight. (iii) how can we have trustworthy data? and (iv) what are the priorities in terms of trustworthy requirements for challenging applications? Regarding this last question, six different applications have been discussed, including trustworthy AI in education, environmental science, 5G‐based IoT networks, robotics for architecture, engineering and construction, financial technology, and healthcare. The review emphasises the need to address trustworthiness in AI systems before their deployment in order to achieve the AI goal for good. An example is provided that demonstrates how trustworthy AI can be employed to eliminate bias in human resources management systems. The insights and recommendations presented in this paper will serve as a valuable guide for AI researchers seeking to achieve trustworthiness in their applications.

Predictive process analytics often apply machine learning to predict the future states of a running business process. However, the internal mechanisms of many existing predictive algorithms are opaque and a human decision-maker is unable to understand why a certain activity was predicted. Recently, counterfactuals have been proposed in the literature to derive human-understandable explanations from predictive models. Current counterfactual approaches consist of finding the minimum feature change that can make a certain prediction flip its outcome. Although many algorithms have been proposed, their application to multi-dimensional sequence data like event logs has not been explored in the literature.In this paper, we explore the use of a recent, popular model-agnostic counterfactual algorithm, DiCE, in the context of predictive process analytics. The analysis reveals that DiCE is unable to derive explanations for process predictions, due to (1) process domain knowledge not being taken into account, (2) long traces of process execution that often tend to be less understandable, and (3) difficulties in optimising the counterfactual search with categorical variables. We design an extension of DiCE, namely DiCE4EL (DiCE for Event Logs), that can generate counterfactual explanations for process prediction, and propose an approach that supports deriving milestone-aware counterfactual explanations at key intermediate stages along process execution to promote interpretability. We apply our approach to a publicly available real-life event log and the analysis results demonstrate the effectiveness of the proposed approach.

The volume of process-related data is growing rapidly: more and more business operations are being supported and monitored by information systems. Industry 4.0 and the corresponding industrial Internet of Things are about to generate new waves of process-related data, next to the abundance of event data already present in enterprise systems. However, organizations often fail to convert such data into strategic and tactical intelligence. This is due to the lack of dedicated technologies that are tailored to effectively manage the information on processes encoded in process models and process execution records. Process-related information is a core organizational asset which requires dedicated analytics to unlock its full potential. This paper proposes a framework for devising process querying methods, i.e., techniques for the (automated) management of repositories of designed and executed processes, as well as models that describe relationships between processes. The framework is composed of generic components that can be configured to create a range of process querying methods. The motivation for the framework stems from use cases in the field of Business Process Management. The design of the framework is informed by and validated via a systematic literature review. The framework structures the state of the art and points to gaps in existing research. Process querying methods need to address these gaps to better support strategic decision-making and provide the next generation of Business Intelligence platforms.

The growth of APIs and Web services on the Internet, especially through larger enterprise systems increasingly being leveraged for Cloud and software-as-a-service opportunities, poses challenges to improving the efficiency of integration with these services. Interfaces of enterprise systems are typically larger, more complex and overloaded, with single operation having multiple data entities and parameter sets, supporting varying requests, and reflecting versioning across different system releases, compared to fine-grained operations of contemporary interfaces. We propose a technique to support the refactoring of service interfaces by deriving business entities and their relationships. In this paper, we focus on the behavioural aspects of service interfaces, aiming to discover the sequential dependencies of operations (otherwise known as protocol extraction) based on the entities and relationships derived. Specifically, we propose heuristics according to these relationships, and in turn, deriving permissible orders in which operations are invoked. As a result of this, service operations can be refactored on business entity CRUD lines, with explicit behavioural protocols as part of an interface definition. This supports flexible service discovery, composition and integration. A prototypical implementation and analysis of existing Web services, including those of commercial logistic systems (Fedex), are used to validate the algorithms proposed through the paper.

Abstract Substantial research efforts have been expended to deal with the complexity of concurrent systems that is inherent to their analysis, e.g., works that tackle the well-known state space explosion problem. Approaches differ in the classes of properties that they are able to suitably check and this is largely a result of the way they balance the trade-off between analysis time and space employed to describe a concurrent system. One interesting class of properties is concerned with behavioral characteristics. These properties are conveniently expressed in terms of computations, or runs , in concurrent systems. This article introduces the theory of untanglings that exploits a particular representation of a collection of runs in a concurrent system. It is shown that a representative untangling of a bounded concurrent system can be constructed that captures all and only the behavior of the system. Representative untanglings strike a unique balance between time and space, yet provide a single model for the convenient extraction of various behavioral properties. Performance measurements in terms of construction time and size of representative untanglings with respect to the original specifications of concurrent systems, conducted on a collection of models from practice, confirm the scalability of the approach. Finally, this article demonstrates practical benefits of using representative untanglings when checking various behavioral properties of concurrent systems.

Several methods for enterprise systems analysis rely on flow-oriented representations of business operations, otherwise known as business process models. The Business Process Modeling Notation (BPMN) is a standard for capturing such models. BPMN models facilitate communication between domain experts and analysts and provide input to software development projects. Meanwhile, there is an emergence of methods for enterprise software development that rely on detailed process definitions that are executed by process engines. These process definitions refine their counterpart BPMN models by introducing data manipulation, application binding, and other implementation details. The de facto standard for defining executable processes is the Business Process Execution Language (BPEL). Accordingly, a standards-based method for developing process-oriented systems is to start with BPMN models and to translate these models into BPEL definitions for subsequent refinement. However, instrumenting this method is challenging because BPMN models and BPEL definitions are structurally very different. Existing techniques for translating BPMN to BPEL only work for limited classes of BPMN models. This article proposes a translation technique that does not impose structural restrictions on the source BPMN model. At the same time, the technique emphasizes the generation of readable (block-structured) BPEL code. An empirical evaluation conducted over a large collection of process models shows that the resulting BPEL definitions are largely block-structured. Beyond its direct relevance in the context of BPMN and BPEL, the technique presented in this article addresses issues that arise when translating from graph-oriented to block-structure flow definition languages.

A service-oriented system is composed of independent software units, namely services, that interact with one another exclusively through message exchanges. The proper functioning of such system depends on whether or not each individual service behaves as the other services expect it to behave. Since services may be developed and operated independently, it is unrealistic to assume that this is always the case. This article addresses the problem of checking and quantifying how much the actual behavior of a service, as recorded in message logs, conforms to the expected behavior as specified in a process model. We consider the case where the expected behavior is defined using the BPEL industry standard (Business Process Execution Language for Web Services). BPEL process definitions are translated into Petri nets and Petri net-based conformance checking techniques are applied to derive two complementary indicators of conformance: fitness and appropriateness . The approach has been implemented in a toolset for business process analysis and mining, namely ProM, and has been tested in an environment comprising multiple Oracle BPEL servers.

The business process modelling notation (BPMN) is a graph-oriented language in which control and action nodes can be connected almost arbitrarily. It is supported by various modelling tools but so far no systems can directly execute BPMN models. The business process execution language for Web services (BPEL) on the other hand is a mainly block-structured language supported by several execution platforms. In the current setting, mapping BPMN models to BPEL code is a necessary step towards unified and standards-based business process development environments. It turns out that this mapping is challenging from a scientific viewpoint as BPMN and BPEL represent two fundamentally different classes of languages. Existing methods for mapping BPMN to BPEL impose limitations on the structure of the source model. This paper proposes a technique that overcomes these limitations. Beyond its direct relevance in the context of BPMN and BPEL, this technique addresses difficult problems that arise generally when translating between flow-based languages with parallelism