ABSTRACT To address the issues of low sampling quality and efficiency in processing large‐scale event logs in existing business process event log sampling methods, a new method, named VariantRank, is proposed, which is based on the importance of trace variants. First, the importance of each trace variant is calculated based on the activity importance and the importance of directly‐follow relationships within the trace variants. Then, the trace variants are ranked according to their importance. Finally, based on the given sampling rate and the ranking of trace variants, the final sampling is performed to obtain the sample event logs. The effectiveness of the proposed sampling method is evaluated in terms of both sampling quality and sampling efficiency across 8 public event log datasets. The experimental analysis shows that, compared with the state‐of‐the‐art sampling methods, VariantRank improves the sampling efficiency while ensuring the sampling quality.

ABSTRACT Business process monitoring involves real‐time supervision of a series of activities carried out by an organization to achieve specific objectives. Process event monitoring points (PEMP) are used to pinpoint specific locations within a process model where expected events are anticipated to occur. However, in general business process modeling languages, such as business process model and notation (BPMN), there is a lack of explicit modeling for PEMPs. This article proposes a method for modeling pairwise event monitoring points in business process models to track the execution status of specific activities or process segments via BPMN extension. Specifically, process monitoring points are designed by expanding the modeling element of sequence flow. Moreover, the business process model is decomposed using the refined process structure tree (RPST) to verify the soundness of the designed pairwise monitoring points. The proposed method allows flexible monitoring at the process segment level. And the process monitoring data could be used for a clear understanding of business progress, especially useful in heterogeneous business process model to support decision‐making. Through case study from real‐world business processes, the effectiveness and usefulness of the proposed process monitoring modeling method is validated.

Connected Autonomous Vehicle (CAV) Driving, as a data-driven intelligent driving technology within the Internet of Vehicles (IoV), presents significant challenges to the efficiency and security of real-time data management. The combination of Web3.0 and edge content caching holds promise in providing low-latency data access for CAVs’ real-time applications. Web3.0 enables the reliable pre-migration of frequently requested content from content providers to edge nodes. However, identifying optimal edge node peers for joint content caching and replacement remains challenging due to the dynamic nature of traffic flow in IoV. Addressing these challenges, this article introduces GAMA-Cache, an innovative edge content caching methodology leveraging Graph Attention Networks (GAT) and Multi-Agent Reinforcement Learning (MARL). GAMA-Cache conceptualizes the cooperative edge content caching issue as a constrained Markov decision process. It employs a MARL technique predicated on cooperation effectiveness to discern optimal caching decisions, with GAT augmenting information extracted from adjacent nodes. A distinct collaborator selection mechanism is also developed to streamline communication between agents, filtering out those with minimal correlations in the vector input to the policy network. Experimental results demonstrate that, in terms of service latency and delivery failure, the GAMA-Cache outperforms other state-of-the-art MARL solutions for edge content caching in IoV.

Business process modeling is widely used in modern organizations for business description. Business Process Modeling Notation (BPMN), as a de-facto modeling standard, represents business process models in graphical notations. Nevertheless, BPMN lacks intuitive modeling tasks for Industrial Internet application scenarios (e.g., IoT tasks and multi-instance tasks with constraints). Although there are some works on extending BPMN elements to improve the model representation, most of them stay in the conceptual model only without tool support, or they are confined to specific domains. In this paper, we extend both BPMN elements and attributes for application in the Industrial Internet context, and two modeling tools are implemented in a client version and Web version to support business process modeling via low-code, enabling the BPMN extension model from conceptual to executable level. Two real-world case studies in Industrial Internet are conducted to show the usefulness of process models with BPMN extension. Furthermore, a comprehensive user experiment is conducted to evaluate the extended process models and tools, and the experimental results show that process models with extension have better quality compared with traditional process models and provided tools are effective for business process modeling in Industrial Internet applicationsNote to Practitioners—This article was motivated by the problem of business process modeling for Industrial Internet. Practically, existing methods extend BPMN elements and add text annotations to enhance the representation of process models. Although these methods facilitate process participants to understand the process models in detail, most of them focus on conceptual modeling and increase the complexity of process models. Moreover, the process models and multi-instance business constraints of business processes involving IoT elements are difficult to be represented by using native BPMN. This article proposes a comprehensive business process modeling language, named BPMN$++$, which extend both new modeling elements for Industrial Internet business requirements and attributes for BPMN multi-instance tasks with the corresponding graphical notations. Then, BPMN$++$model could be executed on the process engine, enabling the model from conceptual to executable level. Finally, comprehensive experiments are conducted for BPMN$++$.

Learning resources in online learning systems typically adhere to uniform formats and settings, lacking flexibility and personalization to meet diverse learning needs and preferences. This inability to meet individualized learning needs and preferences has spurred research interest in personalized learning path recommendations. Many researchers have explored recommending learning path by leveraging user historical learning resource sequence to model personalized characteristics. However, these methods overlook the time information in the learning process and fail to interpret the dynamic shifts in learning preferences during recommendation. Therefore, we propose a method, termed TA-RL, for learning path recommendation, based on time-aware attention mechanism and reinforcement learning. First, we propose a novel time-aware attention mechanism to trace the evolving learning preferences of user, in which attention weights are computed using a context-aware time distance measure and the similarity between history learning resources. Then, we employ a Monte Carlo policy gradient reinforcement learning method to generate learning path recommendation based on learning preferences. We validate the effectiveness of our proposed method by comprehensive experiments on two real-world datasets.

Summary Web service recommendation as an emerging topic attracts increasing attention due to its important practical significance. As the number of available Web services continues to grow, users face the challenge of searching the most suitable services that meet their specific needs. Quality of service (QoS)‐based service recommendation becomes a popular approach to address this issue. However, existing QoS‐based service recommendation methods are inability to effectively capture valuable content and structural information from services. These methods often rely solely on low‐order explicit feature intersections in QoS information, do not fully utilize the high‐order implicit feature intersections, and ignore the rich semantic information existing in service descriptions and user preferences. To address this problem, this paper proposes a Web service recommendation method via combining knowledge distillation representation and DCNMIX quality prediction. This method combines content‐based and structure‐based service classification and service prediction based on multi‐dimensional service quality information. First, it builds a service relationship network using semantic features extracted from service descriptions. Second, it designs a graph neural network knowledge distillation framework. The teacher model extracts the knowledge of the graph neural network model, and the student model learns the structure‐based and feature‐based prior knowledge of the service relationship network. Then the student model is used to learn the knowledge of the teacher model, classify Web services, and obtain service representations. Finally, based on service representations and multi‐dimensional QoS information, it exploits the DCNMIX model to learn the explicit and implicit features intersections of Web services and obtain the prediction score and ranking of Web services. The experimental results on the ProgrammableWeb dataset show that the proposed method outperforms the state‐of‐the‐art baselines in terms of Recall, F1, Logloss, and AUC_ROC.

Abstract As a new paradigm of distributed computing, mobile edge computing (MEC) has gained increasing attention due to its ability to expand the capabilities of centralized cloud computing. In MEC environments, a software application typically consists of multiple micro‐services, which can be composed together in a flexible manner to achieve various user requests. However, the composition of micro‐services in MEC is still a challenging research issue arising from three aspects. Firstly, composite micro‐services constructed by ignoring the processing capabilities of different micro‐services may cause waste of edge resources. Secondly, edge servers' limitations in terms of computational power can easily cause service occupancy between composite micro‐services, severely affecting the user experience. Thirdly, in dynamic and unstable mobile environments, different edge users have different sensitivities to request latency, which increases the complexity of micro‐service composition. In order to improve edge resource utilization and user experience on micro‐service invocations, in this paper, we comprehensively consider the above three factors, and we first model the micro‐services composition problem in MEC as a constrained multi‐objective optimization problem. Then, a micro‐service composition optimization method M3C combining graph search and branch‐and‐bound strategy is proposed to find a composition solution set with low energy consumption and high success rate for multiple edge users. Finally, we perform a series of experiments on two widely used datasets. Experimental results show that our proposed approach significantly outperforms the four competing baseline approaches, and that it is sufficiently efficient for practical deployment.

With the wide adoption of SOA (Service Oriented Architecture) in software engineering, a large number of Web services have emerged to meet the Mashup development requirements. Due to the existence of numerous Web services with similar or identical functionalities, it is challenging for users to select the appropriate Web API for Mashup creation, which makes Web service recommendation an effective approach. The performance of current FM-based service recommendation methods is limited by the sparsity of semantic features related to their functionalities. Furthermore, the network structure features of Web services are often overlooked. However, these features are of great importance and should be incorporated into the service recommendation process. Based on the above considerations, this paper proposes a service recommendation model which fuses content information and network information. Firstly, service content information and network structure information are extracted respectively. Then, these two types of information are characterized separately, and their functional semantics are extracted. Finally, the above information is fused and processed by Neural and Attention Factorization Machine to obtain the final recommendation results. Experimental results show that fusing service network representation information can effectively improve the accuracy of service recommendation results.

With the rapid development of service computing, a large number of methods for Web service recommendation have been proposed. However, the existing approaches using Mashup description information ignore the fact that the users without knowledge of Web APIs are not able to describe their needs in detail, let alone find Web services that meet those needs and are compatible with each other. Meanwhile, most approaches that utilize Web API collaboration network based on Mashup-API invocation relationships do not effectively capture the local and global structure between APIs and mine hidden API compatibility information in the network. This paper introduces the KS-GNN model, a novel approach that utilizes graph neural network and auto-encoder techniques for Web API recommendation. Firstly, we utilize KeyBert to extract keywords related to Web services from functional descriptions. Then, we embed the extracted keywords and use their embedded representations as node representation vectors on the Web API collaboration network. Finally, considering local and global structural relationships in the Web API collaborative network and the network structural relationships for message passing, KS-GNN performs keyword searching on the Web API collaborative network, to recommend the top-K Web services that match the user’s query. Experimental results on the ProgrammableWeb dataset show that KS-GNN outperforms other deep learning-based factorization machine recommendation models. In the meantime, we also confirm that the method of extracting keywords using KeyBert outperforms other keyword extraction methods.

With the explosive growth of data, how to apply these data with time characteristics to recommendation is particularly important.In order to do that, an attention-based time-aware neural collaborative filtering approach named AT-NCF is proposed in this paper.Concretely, short-term preferences of users are learned by a time perception network, and long-term preferences part are modeled by a historical interaction matrix.Then, the short-term preferences are concatenated with long-term preferences as the dynamic user preference vector.After that, high-order user-item feature interactions are learned by a general neural collaborative filtering framework which includes two types of DL models, Deep Matrix Factorization (DMF) and Multiple-Layer Perception (MLP).Finally, the predicted scores are output from the final layer of AT-NCF.Experimental results on the realworld e-commerce dataset verify the effectiveness of the proposed method.

Summary To meet the ever‐increasing requirements of the applications, cloud service providers have further built and managed multiple cloud data centers in multiple regions across geographies with many physical machines (PMs). However, most of the existing resource allocation algorithms are developed for a single cloud data center, which normally cannot efficiently handle the load burst occasions where a single cloud data center may not be enough to satisfy the demands burst of applications. Therefore, it is necessary to consider how to efficiently manage multiple cloud data centers while meeting application requirements and reducing energy consumption. This paper first systematically analyzes multiple cloud data centers and energy consumption models. Then, an energy‐efficient method of Resource Allocation based on Request Prediction in multiple cloud data centers (RARP) is proposed. The RARP method constructs a resource allocation framework based on request prediction in multiple cloud data centers, which anticipates the application request volume in advance. At the same time, the RARP method allocates VMs and PMs based on the principle of minimum remaining resources available to achieve minimum usage of PMs, thus minimizing energy consumption to complete application requests. Extensive experiments are conducted on the proposed RARP method through the simulation platform CloudSim. Finally, the experimental test results show that the accuracy of request detection and the energy consumption of cloud data centers are significantly better than those of the comparison algorithms.

Summary Industrial Internet is a promising technology combining industrial systems with Internet techniques to significantly improve production efficiency and reduce cost by cooperating with intelligent devices. Under industrial internet environment, a production process usually consists of multiple subtasks, and one or more types of product factors are needed to finish a subtask. Thus, the service composition under industrial application is more complex and challenging compared with traditional service composition under the Internet environment. In this article, we model the problem of service composition for production factors under industrial internet as a multiobjective optimization problem. To derive the optimal Pareto service composition plans, we propose a hybrid optimization algorithm, named TLBO‐TS, by combining the advantages of teaching‐learning‐based optimization algorithm and tabu search algorithm. Extensive experiments are conducted to compare with other population‐based optimization methods under a real‐world ship production process to verify the superiority of our approach.

Business process monitoring aims at identifying how well running processes are performing with respect to performance measures and objectives. The existing business process monitoring techniques focus on collecting and analyzing information on the way business processes themselves are executed. They neglect the evaluation of task-roles which play a key role in the performance of the whole business process execution. Different from the traditional perspective, this paper focuses on monitoring the behavior of task-roles and evaluating their performance with respect to timeliness. Specifically, this paper proposes to promote the performance of task-roles by time reminder via business process monitoring, which is implemented by semantic extension of BPMN elements. Further, we extend the information of process execution event log data, with which the performance of task-roles can be evaluated by analyzing the extended event log data. An empirical study of the proposed approach with real-world business processes reveals the effectiveness with respect to performance evaluation of task-roles.

Service classification helps to improve the efficiency of service discovery. Previous methods mainly focus on homogeneous graph-based service classification. However, due to the heterogeneity of service data in the real world, these methods cannot deal with many types of nodes and edges in service relationship network well, and lack the usage of rich semantic information. The emergence of heterogeneous graph attention network can effectively solve the problems, because it can more completely and naturally extracts the relationships and nodes from the service relationship network, and well distinguishes the importance of neighbor nodes and meta paths. Therefore, this paper proposes a heterogeneous graph attention network-enhanced Web service classification method. In this method, firstly, a heterogeneous information service network is constructed by using composite service information, atomic service information and their attribute information. Then, the meta path is defined according to different semantic information, and the similarity matrix of service is constructed by using the commuting matrix and the similarity measurement technology based on meta path. Finally, a two-layer attention model is designed to calculate the node-level attention and meta path-level attention of the service, so as to obtain the node-level representations and meta path-level representations of the services, and generate more representative embedding features of services for achieving more accurate service classification. Finally, the experimental results on real datasets of ProgrammableWeb show that our method is better than GAT, GCN, Metapath2Vec, Node2Vec, BiLSTM and LDA in terms of precision, recall and macro F1, and improves the accuracy of Web service classification.

Summary Traditional market basket recommendation approaches normally cannot well recommend unpopular commodities in big data environment. To address such problem and deal with large datasets of practical supermarkets, this paper presents a market basket recommendation framework and proposes an Extended algorithm based on Collaborative Filtering and Association Rule mining, named ECFAR. The ECFAR covers two sub‐algorithms. First, a parallel FP‐Growth algorithm is used for mining association rules on Spark, which is designed to increase the efficiency of processing big data. Then, a parallel similar commodity discovery method based on matrix factorization is proposed. By analyzing a real‐world sales dataset collected from a local supermarket group, extensive experiments are conducted to verify its effectiveness.

Summary Resource usage prediction is increasingly important in cloud computing environments, and CPU usage prediction is especially helpful for improving the efficiency of resource provisioning and reducing energy consumption of cloud datacenters. However, accurate CPU usage prediction remains a challenge and few works have been done on predicting CPU usage of physical machines in cloud datacenters. In this article, we present a deep belief network (DBN) and particle swarm optimization (PSO) based CPU usage prediction algorithm, which is named DP‐CUPA and aimed to provide more accurate prediction results. The DP‐CUPA consists of three main steps. First, the historic data on CPU usage are preprocessed and normalized. Then, the autoregressive model and grey model are adopted as base prediction models and trained to provide extra input information for training DBN. Finally, the PSO is used to estimate DBN parameters and the DBN neural network is trained to predict CPU usage. The effectiveness of the DP‐CUPA is evaluated by extensive experiments with a real‐world dataset of Google cluster usage trace.

Summary Recommender systems are becoming increasingly critical to the success of commerce sales. In spite of their benefits, they suffer from some major challenges including recommendation quality such as the accuracy, diversity, and novelty of recommendations. In the context of retail business, the novelty of recommendations is of especial importance because it can directly affect customers' probabilities of buying commodity and whether to visit stores again. However, tradition algorithms for retail commodity recommendation never consider the problem of improving the novelty of recommendations. To address this, a novel multiarmed bandit and gradient boosting decision tree‐based retail commodity recommendation approach is proposed in this article, which is named MGRCR. It can increase recommendations' novelty while maintaining comparable levels of in the context of retailing. The effectiveness of our proposed approach has been proved by comprehensive experiments with real‐world commerce datasets and different state‐of‐the‐art recommendation techniques.

Online learning is becoming increasingly popular in recent years. Personalized recommendation is particularly important for the development of online learning systems. Though LSTM model has been widely applied in various recommendations, it normally can't deal with the problem of sparse data. In this paper, we present a novel model for learning resource recommendation, named G-LSTM. Our model integrates the Generalized Matrix Factorization (GMF) with Long Short-Term Memory (LSTM) model. For evaluating our model, we prepare and analyze two datasets from Junyi Academy. Extensive experiments are conducted on the two datasets to verify the superiority of our model in both effectiveness and accuracy.

Collaborative filtering approach based on rating is one of the most broadly used service recommendation approach. However, rating data is very sparse in most service recommender systems, which seriously impacts the precision of service recommendation. In view of this problem, a usage-based service recommendation approach is proposed in this paper. What is special about this approach is that usage information instead of rating data is recruited to infer user interest. Some experiments are implemented to verify the efficient of this approach.

SUMMARY The employment of batch processing in workflow is to model and schedule activity instances in multiple workflow cases of the same workflow type to optimize business processes execution dynamically. Although our previous works have preliminarily investigated its model and implementation, it is still necessary to deal with its model design problem. Process mining techniques allow for the automated discovery of process models from event logs and have received notable attentions in researches recently. Following these researches, this paper proposes an approach to mine batch processing workflow models from event logs by considering the batch processing relations among activity instances in multiple workflow cases. The notion of batch processing feature and its corresponding mining algorithm are also presented for discovering the batch processing area in the model by using the input and output data information of activity instances in events. The algorithms presented in this paper can help to enhance the applicability of existing process mining approaches and broaden the process mining spectrum. Copyright © 2013 John Wiley & Sons, Ltd.

Activity instance oriented handling is a new means for vertical optimization of process cases. Unlike our previous batch processing mechanism in workflows, it focuses on the data characteristics of activity instances and utilizes explicit knowledge for execution optimization. This paper introduces its concept and investigates its modeling and enactment mechanisms. It uses activity instance pattern as a base to model and represent the knowledge for execution optimization. The system design for activity instance oriented handling and related algorithms are also proposed.

Abstract Batch processing in workflow schedules activity instances in multiple workflow cases of the same workflow type to run as a group. It can optimize business processes execution dynamically. To achieve this goal, it is necessary to define a dataflow operation language to group and ungroup the data in multiple cases of a workflow. Though our previous work has preliminarily investigated the model and its implementation, there is still lack of a formally defined model. In this paper, we first propose a method that is based on a partial vector space to model the dataflow in multiple workflow cases. Based on this model, the data operation primitives for batch processing are specified and defined formally. Since most WfMSs (Workflow Management Systems) use RDBMS (relational database management system) to store their data currently, an SQL (Structured Query Language)‐like implementation language, namely DBOL (Data Batch Operation Language), is proposed. Evaluation experiments have also been done to show its performance. Copyright © 2011 John Wiley & Sons, Ltd.

The instance-aspect oriented workflow management system is to vertically combine multiple workflow activity instances and submit them for execution as a whole according to some batch or combination logics. It is inspired by the idea of aspect-oriented programming methodology and aims at improving the execution efficiency of business processes. Traditional workflow systems do not support workflow model with instance aspects. In our previous work, we have studied workflow instance modeling technology. This paper makes a research on the principles, methods and implementation of a workflow visual GUI tool for modeling instance aspects in workflow. It is based on an open source GUI tool, Together Workflow Editor, and makes some expansion in instance aspect functionality.