ABSTRACT Predictive process monitoring is a log analysis technique that combines data and processes. It predicts the future behavior of ongoing cases based on historical process execution records, thereby supporting rational resource allocation and intelligent process decision making. Existing methods based on deep learning techniques mainly rely on single‐case trends to predict the next activity or remaining time, neglecting the impact of global structural context on the prediction of currently ongoing cases. To address this limitation, this paper proposes a Predictive Business Process Monitoring framework based on Graph contrastive Learning and Attention Mechanism (BPMGLAM). The framework focuses on both direct and indirect relationships between activities to construct a global graph structure. Through a contrastive learning strategy between global graph‐related subgraphs and local graphs, it enhances the structural representation capability of ongoing cases and integrates multi‐level event attribute information to generate reliable predictions of future activities. Experiments conducted on multiple real‐world event logs demonstrate that the proposed method achieves superior prediction performance.

Existing Predictive Process Monitoring (PPM) methods typically rely on static offline methods, limiting their ability to adapt to dynamic process evolution driven by emerging activities and shifting behavioral patterns. This constraint is particularly critical for multi-task prediction, such as the simultaneous forecasting of the next activity and the remaining process time. To address this challenge, we propose an online multi-task prediction framework based on dynamic graph representations. The framework enables a Graph Neural Network (GNN) to incrementally learn newly emerging activities by leveraging dynamic graph snapshots and an architecture expansion strategy. For efficient online adaptation, the framework incorporates two update strategies, a standard periodic update and a drift-aware adaptive update triggered by the Maximum Mean Discrepancy (MMD2) between subgraph embeddings. Both strategies are integrated with a Prioritized Experience Replay (PER) mechanism, augmented with a rarity-aware bonus, to ensure rapid and robust model adjustments in non-stationary environments. Comprehensive experiments on multiple real-world event logs demonstrate that our framework, when combined with different GNN backbones such as GCN, GAT, and GIN, significantly outperforms state-of-the-art baselines in both next-activity and remaining time prediction. Notably, under concept drift, the proposed drift-aware strategy exhibits strong adaptability, highlighting the framework’s effectiveness and potential for addressing complex online process prediction challenges.

Predictive Process Monitoring (PPM) aims to predict the future states of ongoing process instances. A primary objective is to accurately predict process outcomes while ensuring decision transparency, which is critical for enhancing process efficiency and reducing operational risk. Existing interpretable approaches to process monitoring often struggle with balancing transparency and reliability. Specifically, approaches that prioritize transparency often fall short in predictive accuracy and generalization, while those that achieve higher prediction performance often provide less reliable explanations. To address these limitations, we propose a novel Transparent Process Outcome Prediction framework (TPOP) using a graph neural network with stochastic attention. We begin by applying a SHAP-based feature selection technique to identify and extract the most relevant attributes from the log, thereby improving the quality of graph-based process representations. Next, we introduce a graph stochastic attention mechanism, which helps the model in concentrate on key paths and activities during training, leading to transparent and trustworthy predictions. Experimental evaluations on ten real-life event logs demonstrate that our approach outperforms state-of-the-art approaches in both predictive performance and interpretability. Furthermore, by visualizing how specific activities influence process outcomes across various cases, we confirm the reliability of the explanations generated by our approach.

ABSTRACT Process discovery aims to construct business process models from event data recorded by information systems. Traditional integer linear programming (ILP)‐based process discovery methods can extract complex local and nonlocal process structures. However, they often generate overfitted models when dealing with large‐scale logs, and ILP solvers are computationally inefficient. Moreover, event logs inevitably contain noise or outliers, and existing ILP methods struggle to distinguish between effective infrequent behaviors and noise. To address these challenges, this paper proposes a heuristic‐enhanced ILP‐based process discovery method that integrates region‐based constraint formulation and a dynamic threshold mechanism guided by multidimensional dependency metrics for effective noise filtering and model optimization. The method begins with heuristic preprocessing of the event logs, followed by constructing and solving region‐based ILP constraints to obtain an initial Petri net process model. Next, model structure replay and ILP are applied for preliminary noise filtering, and multidimensional dependency metrics are used to compute comprehensive dependency scores. Dynamic thresholds are then applied to further filter noise, enabling iterative optimization of the Petri net model. Finally, experimental evaluations on real and synthetic event logs demonstrate that the proposed method not only preserves effective infrequent behaviors but also significantly improves the quality of the discovered process models.

ABSTRACT Unexpected behavior in business process executions can be identified to give industrial internet systems security assurance for reliable operation. Current research primarily employs consistency analysis or outlier detection of data points to recognize aberrant behavior, neglecting the relationship between behavior and data properties. This work presents a multiperspective association rule‐based approach for detecting anomalous behavior in industrial processes. Initially, a log transaction table with behavior relationships is constructed by mining behavior associations and related properties from the data Petri net. Subsequently, through the application of context awareness, behavior‐attribute‐time associations of frequently occurring itemsets are generated, and pruning procedures are used to mine multiperspective behavior rules under attribute associations. This approach facilitates the identification of anomalous behavior by comparing the support between logs and rules. Ultimately, the proposed method is implemented using the pm4py open‐source framework, and evaluations are performed on both simulated and real event logs using multiple metrics. Experimental comparison results demonstrate that the proposed anomaly behavior detection method achieves higher performance.

Abstract Recently, fake news detection on social media (SM) has attracted a lot of attention. With the emergence of fake news at a breakneck pace, the massive spread of fake news has had a serious impact in our society. The authenticity of the news is questionable and there exists a necessity for an automated tool for the detection. However, most fake news detection methods are mainly supervised, requiring huge amounts of annotated data, which is time-consuming, expensive, and almost impossible with vast new SM volume. To deal with this problem, in this paper, we propose a novel unsupervised fake news detection framework based on structural contrastive learning by combining the propagation structure of news and contrastive learning to achieve unsupervised training. To validate the influence of parameters and our method’s performance, we design experiment sets on public Twitter and Weibo datasets, which validate our approach outperforms current baseline ones and has proper robustness.

Abstract Malicious sellers frequently collaborate with spammers to fabricate reviews for promoting their products. These spammers act strategically in groups and have even formed black-and-gray industry chains. Researchers have proposed Frequent Item Mining-based, review burst-based, and graph-based schemes for spammer group detection. However, existing graph-based schemes often model reviewer relationships as a homogeneous network, failing to fully utilize the relationship semantics between reviewers or account for the burst characteristics of reviews, resulting in poor detection performance. Thus, this research proposes a cross-view contrastive learning-based spammer group detection algorithm for heterogeneous networks. To mine and embed the burst characteristics of reviews, we first filter out the target products and mine the active sessions of reviews about these products. We then construct a heterogeneous network featuring four node types and devise three unique meta-paths. Additionally, we utilize the cross-view contrastive learning method to learn the reviewer embeddings and apply DBSCAN to identify suspected groups for subsequent cleansing and ranking, ultimately determining potential spammer groups. Our experiments show the detection performance of the proposed scheme outperforms baseline ones.

Fueled by advancements in intelligent transportation systems, the Internet of Vehicles (IoV) seeks to connect smart vehicles, road infrastructure, and users into a unified network, enhancing traffic efficiency and reducing accident risks. Centralized cloud data collection raises concerns about privacy and communication overhead. To address these, distributed edge intelligence (DEI) reduces transmission costs and improves privacy by implementing machine learning at the network edge. In this context, cooperative adaptive cruise control (CACC) systems, combined with DEI in the IoV framework, enhance transportation system intelligence through real-time data processing and decentralized decision making. This article proposes a modeling and analysis method for CACC systems based on Petri nets. The datasets are automatically generated using tools developed by our team, and machine-learning methods are utilized to perform risk prediction analysis on the CACC model. From the perspective of Petri nets synchronization, we propose risk mitigation strategies from a design standpoint. The research results show that the proposed method significantly reduces signal accumulation and enhances synchronization in CACC systems. This improvement provides new theoretical support and technical guidance for the design and implementation of CACC systems, ultimately enhancing their safety and reliability.

Abstract Within the thriving e-commerce landscape, some unscrupulous merchants hire spammer groups to post misleading reviews or ratings, aiming to manipulate public perception and disrupt fair market competition. This phenomenon has prompted a heightened research focus on spammer groups detection. In the e-commerce domain, current spammer group detection algorithms can be classified into three categories, i.e., Frequent Item Mining-based, graph-based, and burst-based algorithms. However, existing graph-based algorithms have limitations in that they did not adequately consider the redundant relationships within co-review graphs and neglected to detect overlapping members within spammer groups. To address these issues, we introduce an overlapping spammer group detection algorithm based on deep reinforcement learning named DRL-OSG. First, the algorithm filters out highly suspicious products and gets the set of reviewers who have reviewed these products. Secondly, taking these reviewers as nodes and their co-reviewing relationships as edges, we construct a homogeneous co-reviewing graph. Thirdly, to efficiently identify and handle the redundant relationships that are accidentally formed between ordinary users and spammer group members, we propose the Auto-Sim algorithm, which is a specifically tailored algorithm for dynamic optimization of the co-reviewing graph, allowing for adjustments to the reviewers’ relationship network within the graph. Finally, candidate spammer groups are discovered by using the Ego-Splitting overlapping clustering algorithm, allowing overlapping members to exist in these groups. Then, these groups are refined and ranked to derive the final list of spammer groups. Experimental results based on real-life datasets show that our proposed DRL-OSG algorithm performs better than the baseline algorithms in Precision.

Summary The current process mining method takes high‐frequency behavior as the mainstream behavior, and directly filters out the infrequent logs as noise to obtain a concise business process model. However, effective infrequency behaviors that are important to business processes are often data constrained. From a control flow perspective, it is difficult to accurately capture the effective infrequency behavior. A method for mining effective infrequent behaviors based on data attributes is proposed to solve the above problems. First, the important data attributes of target business processes are obtained by feature combination. Then, attribute assignment rules are set according to the needs of the business process to determine whether it has a beneficial impact on the business process. Lastly, it is suggested that a confidence interval be used instead of the traditional threshold to evaluate and mine effective low‐frequency behavior. The experiment results show that compared with other methods, it can significantly improve the fitness of the business process model and can more accurately mine effective infrequency behavior to optimize the business process model.

Model incremental updating enhances the initial model by analysing discrepancy parts of the system to improve the model's adaptability for new scenarios. These discrepancy components originate from the deviation between the increasing business process operational state and the outdated original planning model. Considerable domain-specific knowledge is required to determine the threshold points for selecting appropriate activities, but process analysts rarely know each scenario's domain knowledge. Moreover, most analytical processes only focussed on the control flow level. Aiming at these issues, this paper proposes a Hybrid Behavioural and Resource Trusted intervals Updating algorithm (BI&RI Updating) for model enhancement based on control level and resource level of online event streams. First, analyse the reference model to construct multi-perspective trusted interval constraints in the offline stage. From a control-flow perspective, the behavioural relationships between activities are researched using a deep clustering approach. Moreover, from a data-flow standpoint, resource co-occurrence relationships are analysed based on association rules. Next, the incremental update algorithm in online scenarios is proposed to update the model by filtering the event streams and iteratively optimising the trusted intervals. Finally, the proposed algorithm is implemented based on the PM4PY framework and evaluated using real logs with both outcome model quality and execution efficiency. The results show that the algorithm can execute quickly and improve the quality of the model even in a non-ideal state where the logs contain noise.

Process discovery usually analyses frequent behaviour in event logs to gain an intuitive understanding of processes. However, there are some effective infrequent behaviours that help to improve business processes in real life. Most existing studies either ignore them or treat them as harmful behaviours. To distinguish effective infrequent sequences from noisy activities, this paper proposes an algorithm to analyse the distribution states of activities and the strong transfer relationships between behaviours based on maximum probability paths. The algorithm divides episodic traces into two categories: harmful and useful episodes, namely noisy activities and effective sequences. First, using conditional probability entropy, the infrequent logs are pre-processed to remove individual noisy activities that are extremely irregularly distributed in the traces. Effective sequences are then extracted from the logs based on the state transfer information of the activities. The algorithm is based on a PM4Py implementation and is validated using synthetic and real logs. From the results, the algorithm not only preserves the key structure of the model and reduces noise activity, but also improves the quality of the model.

Business process recommendation can be used to simplify the working procedures of enterprises, avoid unnecessary expenses, and promote the development of enterprises. In the process of process recommendation, there are a lot of activities that are similar in structure and difficult to choose. Here, a process recommendation method based on cost constraints is proposed to solve the problem of difficult to distinguish similar processes. First, the business process is transformed into a labelled Petri net, and the execution probability of each transition is calculated according to the business process log. Then, the matrix used to represent Petri nets is constructed according to the adjacent relationship between transitions, and the matrix is made into the same dimension, and the similarity between matrices is calculated by biggest–smallest approach degree, and the set of Petri nets with similar structure is established. Finally, a cost constraint-based process recommendation method is proposed to find lower service cost items in similar process sets. In the experimental part, the feasibility of the method is compared and verified.

Mixing patterns (MPs) in social trust networks (STNs) are increasingly attracting attention because they can assist analysts in designing information dissemination tactics and planning electronic word-of-mouth (eWOM) campaigns. However, the existing studies on MPs do not explain the assortative or disassortative tendencies of STNs due to their omission of the support of the sociological theory, as well as that of network theory. To address this issue, this study investigates the MPs in STNs from the standpoint of social identity theory (SIT). The user trust networks (UTNs) are modeled by a directed multigraph (DMG). Then, the structural properties of homogeneous trust networks and heterogeneous trust networks are explored via measures that include degree centrality, the correlation coefficient (CC), the cumulative distribution of the ratio of trust degree to distrust degree (CDRTD), and the assortativity coefficient. The MPs of homogeneous trust networks and heterogeneous trust networks are explained from the perspective of SIT. An experiential evaluation is conducted in the constructed homogeneous trust networks and heterogeneous trust networks using a real-world data set crawled from Epinions. The research findings indicate that the MPs in homogeneous trust networks tend toward assortative mixing (AM), and those in heterogeneous trust networks tend toward disassortative mixing (DM). The experimental results show that the performance of the proposed approach is superior to that of the state-of-the-art approach to influential user identification.