The Internet of Things, privacy, and technical constraints increase the demand for edge-based data-driven services, which is one of the major goals of Industry 4.0 and Society 5.0. Big data analysis (BDA) is the preferred approach to unleash hidden knowledge. However, BDA consumes excessive resources and time. These limitations hamper the meaningful adoption of BDA, especially the time and situation critical edge use cases, and hinder the goals of Industry 4.0 and Society 5.0. Automating the BDA process at the edge is a cognitive approach to address the aforementioned concerns. Data science workflow is an indispensable challenge for successful automation. Therefore, we conducted a systematic literature survey on data science workflow platforms as the first contribution. Moreover, we learned that the BDA workflow depends on diversified constraints and undergoes rigorous data-mining stages. These caused an increase in the solution space, dynamic constraints, complexity issues, and NP-hardness of BDA workflow. Graphplan is a heuristic AI-planning technique that can address concerns associated with BDA workflow. Therefore, as the second contribution, we adopted the graphplan to generate a workflow for edge-based BDA automation. Experiments demonstrate that the proposed method achieved our objectives.

Raw medical documents come with in-domain challenges that need to be addressed with appropriate methods, such as a suitable tokenizer and tailored model via hyperparameters. Most available pre-trained models have been trained on clean biomedical data from published documents without considering the characteristics of raw medical texts. This approach drastically limits the performance of those models on real-world data such as electronic health record documents. This work introduces a fine-grained language model trained over a SentencePiece tokenizer with various biomedical and clinical data. Through this RoBERTa-based model, referred to as BioBERTa, we demonstrated the importance of data-dependant optimization by evaluating our model on several Named Entity Recognition tasks. Our approach showed state-of-the-art results on NCBI-disease and BC5-disease datasets. We demonstrated the effectiveness of an in-domain tokenizer where ours improved the embedding length by 15.1% of its original model. Our approach also shows that fine-tuning the hyperparameters can significantly improve the model accuracy(up to 2% F1 score).

In Intensive Care Unit (ICU), it is important to anticipate interventions for patients at a high-risk of death. This requires identifying those patients ideally at the time of their admission in ICU, and update their initial risk rate every time new data is available. This predictive task can be performed by analyzing structured and unstructured routine data to make sure that we can initiate a prediction for every patient. Traditional statistical tools have been used to assess disease like pneumonia and predict the outcome of a patient. Recently, machine learning models emerged and have shown better performances on such tasks. Although authors have published various results, their works rely on a single datatype, either structured or unstructured data. Using the Medical Information Mart for intensive Care data-set, we are proposing an ensemble model, that aggregates different data-types to predict the outcome of a pneumonia patient admitted in ICU using limited data that can be available at the very early stage of his stay. To demonstrate the importance of this approach, we compared it with 2 different models, one based only on structured data, and another one based on narratives text from caregivers, and we were able to show that our ensemble model can perform way better with an accuracy of 0.98 of F1-score(0.97 MCC), while a model using only structured data had 0.79 of F1-score and where text notes predicted the outcome with an accuracy of 0.89 of Matthews Correlation Coefficient. In addition to showing how ensemble learning models can outperform other models on this task, we demonstrated the importance and usefulness of interpreting the predictions pointing out the leading factors that are determining the global and individual outcome predictions.

Finding Spatial High Utility Itemsets (SHUIs) in a spatiotemporal database is a challenging problem of great importance in many real-world applications. Most previous works focused on the sequential discovery of SHUIs in a database running on a single machine. Consequently, these works are not suitable for big data (or cloud-based) applications as they suffer from the scalability and fault tolerant problems. This paper proposes several novel pruning techniques to reduce the search space and present a more flexible distributed algorithm to find all desired itemsets from the database using Spark in-memory computing architecture. Our algorithm inherits several advantages of Spark, including low communication cost, fault tolerance, and high scalability. Experimental results demonstrate that the proposed algorithm has good scalability and performance on very large databases. Finally, we present a real-world navigation application in which SHUIs generated from the traffic congestion data have been employed to recommend alternative routes to the users.

Automation of Big Data Analysis (BDA) procedure gives us a great profit in the era of Big Data and Artificial Intelligence. BDA procedure can be efficiently automated by the automatic service composition concept efficiently. Our previous work for Auto-BDA shows a great future prospect in reducing turnaround time for data analysis. Moreover, it requires consideration of the automation with a well-geared combination of the data preparation and the optimal model (deep learning) generation. This paper shows the construction of automating BDA and model generation (here deep learning) together with data preparation and parameter optimization.

A novel framework and method to generate automatically deep learning (DL) services for developers who are not artificial intelligence (AI) experts is presented. Two issues have been considered for the framework: 1) separation of the knowledge on DL from DL generation engines, and 2) cooperation between Automatic Service Composition (ASC) and the CRoss-Industry Standard Process for Data Mining (CRISP-DM) procedure. First, the separation of knowledge on DL from DL generation engines is necessary to adapt to advances and changes in DL technology. Ontology and rules for knowledge and experience in regard to DL technology applications will therefore be constructed for DL generation engines. Second, a framework to compose a target service for generation of a DL architecture requested by non-AI domain experts will be constructed using ASC that works with input involving user requirements based on the CRISP-DM procedure. The created ontology, rules, and composition procedure based on a scenario for special document classification are explained.

The cost-minimization problem for streaming workflow (SW) has already become increasingly important and even critical in stream big data processing, particularly for geographically distributed datacenters, because of its huge demand on computing and communicating resources. Existing virtual machine (VM) allocation algorithms in cloud computing have been widely applied to batch-processing models; however, none of them can be successfully applied to SW because: 1) they failed to adapt the continuous execution characteristic of SW; and 2) most of them are all based on the assumption that the price of traffic and VMs among datacenters are uniform. In this paper, we propose a transformation-based SW allocation algorithm with the goal of cost-minimization for stream big data processing in geographically distributed datacenters, considering the characteristics of SW and price heterogeneity among geographically distributed datacenters. We first propose a cost-aware workflow transformation framework based on eight well-designed and verified transformation rules for cost reduction to adapt the continuous execution characteristic of SW. We then formulate the joint VM-traffic optimization problem and show that it is NP-hard. To produce the optimal solution in polynomial time, we then transform the SW allocation problem into the minimum-cost maximum-flow problem, considering both traffic and VMs price heterogeneity. Finally, our experimental results validate the high cost efficiency of our approach with lower computing and communicating costs by optimizing the workflow specification and joint VM-traffic cost optimization.

Web service recommendation in an efficient and accurate manner has become a significant tool with information overload and an increasingly urgent demand to provide appropriate recommendations to users. Among the service recommendation algorithms, Collaborative Filtering (CF) gives credence to user inputs by comparing user's correlations. Performance of the service recommendation approaches becomes deficient due to the data sparsity and cold-start issues, which make the incomplete and inadequate information to analyze a user predicament on Web services. This paper proposes a CF-based recommendation approach that first alleviates the sparsity problem using a novel ontology-based clustering approach that used domain specificity and service similarity for the ontology generation. Then, we propose a trustbased user rating prediction by determining the trust value between users by calculating the correlation of users. The experimental results indicate that the proposed approach can effectively alleviate the sparsity and cold-start problems by lower prediction error compared with existing sparsity managing mechanisms in service recommendations.

In recent years, due to the growth of information on the internet, the number of available Web services has increased. Clustering Web services based on their functional features to different domains have started to play a major role in several service management tasks such as efficient Web service discovery and recommendations. In this paper, we propose a novel ontology-based approach for Web service clustering. Instead of using traditional methods, we focus on the similarity and specificity of terms for ontology generation. The amount of domain-specific information included in a term is used to define the specificity of that term. Specific terms are more powerful than general terms for describing a large amount of domain information. Taking advantage of this, we generate a new ontology, which is then used to calculate similarity by defining new logic-based filters. When the similarity calculation fails, we apply information retrieval-based methods. Based on a comprehensive evaluation that we conducted to measure the performance of our method, our novel clustering approach was shown to be more effective in terms of precision, recall, Fmeasure, purity and entropy than other existing clustering approaches.

The use of the big data analytics (BDA) platform is increasingly becoming prevalent in the data sciences. However, BDA processes consume resources and time excessively. Automating BDA processes is a cognitive approach to the BDA domain, which is most impaired by its heavy consumption of time and resources. However, the BDA workflow is highly dependent on diversified constraints because of the high variability, veracity, and volume of data processing to accomplish highly influential and sophisticated requirements. The workflow has to pass rigorous and diverse data mining steps, each step contains several tasks and these tasks made by many sub tasks, that it must be accomplished to progress to the next step in the workflow. This increases the available solution space for BDA processes. The intelligent heuristic approach is needed to address the domain-specific concerns which are large solution space, and awareness of constraints are caused the BDA planning. Therefore, we propose to use GraphPlan-based dynamic workflow generation for the BDA domain. Our empirical studies prove that the proposed sophisticated method satisfied planning requirements and outperformed a related planning technique.

The virtual machine (VM) allocation problem in cloud computing has been widely studied in recent years, and many algorithms have been proposed in the literature. Most of them have been successfully applied to batch processing models such as MapReduce; however, none of them can be applied to streaming workflow well because of the following weaknesses: 1) failure to capture the characteristics of tasks in streaming workflow for the short life cycle of data streams; 2) most algorithms are based on the assumptions that the price of VMs and traffic among data centers (DCs) are static and fixed. In this paper, we propose a streaming workflow allocation algorithm that takes into consideration the characteristics of streaming work and the price diversity among geo-distributed DCs, to further achieve the goal of cost minimization for streaming big data processing. First, we construct an extended streaming workflow graph (ESWG) based on the task semantics of streaming workflow and the price diversity of geo-distributed DCs, and the streaming workflow allocation problem is formulated into mixed integer linear programming based on the ESWG. Second, we propose two heuristic algorithms to reduce the computational space based on task combination and DC combination in order to meet the strict latency requirement. Finally, our experimental results demonstrate significant performance gains with lower total cost and execution time.

In the era of Big Data, data analysis gives strong competition power to enterprises. As services for Big Data Analysis (BDA) become prevalent, analysis services with intelligence and autonomy using automatic service composition show very bright prospects in the BDA market. Service composition consists of four stages: workflow generation, discovery, selection, and execution. In this paper, we propose a novel service discovery approach that considers two key concerns in the discovery domain towards better quality as well as effective service composition. BDA services are fine grained according to the domain and functional behaviors. The services need a domain context-aware and precision-guided discovery approach. Therefore, we propose domain ontology-based service discovery. It is mainly focused on the BDA domain for precise service discovery considering all behavioral signatures between queries and services. As for the second concern, components in composed services depend greatly on each other in situations such as workflow for data analysis. We show that linking services together considering sociability or user preference gives better discovery performance. We propose a Linked Social Service Network (LSSN) with multiple feature attribute-based service discovery for BDA. Our approach combines two advantages, the precision and sociability of Web services. The experimental results show that both of these methods perform well based on their perspectives, better than previous approaches.

Nowadays, owing to the growth of quantity of data, the data mining techniques have been required on web exceedingly for extracting information from the data. Classification of text in data mining is very important and has been a hot issue on the topic. Especially, ontological taxonomy classification is important for more intelligent information reasoning. As it relates to data distribution of classes directly, we investigate relation between the data distribution and classification performance in this research. This paper shows the clue to improve taxonomy classification accuracy from a new viewpoint.

We propose a new optimal data placement technique to improve the performance of MapReduce in cloud data centers by considering not only the data locality but also the global data access costs. We first conducted an analytical and experimental study to identify the performance issues of MapReduce in data centers and to show that MapReduce tasks that are involved in unexpected remote data access have much greater communication costs and execution time, and can significantly deteriorate the overall performance. Next, we formulated the problem of optimal data placement and proposed a generative model to minimize global data access cost in data centers and showed that the optimal data placement problem is NP-hard. To solve the optimal data placement problem, we propose a topology-aware heuristic algorithm by first constructing a replica-balanced distribution tree for the abstract tree structure, and then building a replica-similarity distribution tree for detail tree construction, to construct an optimal replica distribution tree. The experimental results demonstrated that our optimal data placement approach can improve the performance of MapReduce with lower communication and computation costs by effectively minimizing global data access costs, more specifically reducing unexpected remote data access.

Big Data analytics provide support for decision making by discovering patterns and other useful information from large set of data. Organizations utilizing advanced analytics techniques to gain real value from Big Data will grow faster than their competitors and seize new opportunities. Cross-Industry Standard Process for Data Mining (CRISP-DM) is an industry-proven way to build predictive analytics models across the enterprise. However, the manual process in CRISP-DM hinders faster decision making on real-time application for efficient data analysis. In this paper, we present an approach to automate the process using Automatic Service Composition (ASC). Focusing on the planning stage of ASC, we propose an ontology-based workflow generation method to automate the CRISP-DM process. Ontology and rules are designed to infer workflow for data analytics process according to the properties of the datasets as well as user needs. Empirical study of our prototyping system has proved the efficiency of our workflow generation method.

A Web service is defined as an autonomous unit of application logic that provides either some business functionality or information to other applications through an Internet connection. Web services are based on a set of eXtensible Markup Language (XML) standards such as Universal Description, Discovery and Integration (UDDI), Web Services Description Language (WSDL), and Simple Object Access Protocol (SOAP). Nowadays Web services are becoming more and more popular for supporting different social applications, thus there are also increasing demands and discussions about Web services privacy protection in information. In general, privacy policies describe an organization's data practices on what information they collect from individuals (e.g., consumers) and what (e.g., purposes) they do with it. To enable privacy protection for Web service consumers across multiple domains and services, the World Wide Web Consortium (W3C) published a document called “Web Services Architecture (WSA) Requirements” that defines some specific privacy requirements for Web services as a future research topic. This paper presents a mathematical model to construct the privacy policies in SOAP Message Exchange Patterns (MEP) for social services. Further, this paper also presents the privacy policies in security tokens with SOAP messages.

Automatic service composition can create new value-added services dynamically and automatically from existing services in an envisioned service-oriented architecture. However, despite considerable progress, web-scale uptake has been significantly less than initially anticipated because of several challenging issues, such as poor scalability, exponentially expanding search time in large search spaces, and the lack of service sociability caused by the isolation of services. In this paper, we propose an innovative methodology for moving from isolated service islands to a global social service network (GSSN) by developing a network model that supports service sociability. First, we propose the construction of a GSSN based on the quality of social links. We then propose an algorithm that maps the GSSN into a service cluster network to reduce the search space, and a quality-driven composition approach that enables exploitation of the service cluster network by providing workflow as a service. Finally, experimental results show that our GSSN-based approach can solve the service composition problem well, improving not only the response time but also the success rate.

Web services have had a tremendous impact on the Web for supporting a distributed service-based economy on a global scale. However, despite the outstanding progress, their uptake on a Web scale has been significantly less than initially anticipated. The isolation of services and the lack of social relationships among related services have been identified as reasons for the poor uptake. In this paper, we propose connecting the isolated service islands into a global social service network to enhance the services' sociability on a global scale. First, we propose linked social service-specific principles based on linked data principles for publishing services on the open Web as linked social services. Then, we suggest a new framework for constructing the global social service network following linked social service-specific principles based on complex network theories. Next, an approach is proposed to enable the exploitation of the global social service network, providing Linked Social Services as a Service. Finally, experimental results show that our approach can solve the quality of service discovery problem, improving both the service discovering time and the success rate by exploring service-to-service based on the global social service network.

Clustering Web services into functionally similar clusters is a very efficient approach to service discovery. A principal issue for clustering is computing the semantic similarity between services. Current approaches use similarity-distance measurement methods such as keyword, information-retrieval or ontology based methods. These approaches have problems that include discovering semantic characteristics, loss of semantic information and a shortage of high-quality ontologies. Further, current clustering approaches are considered only have simple data types in services' input and output. However, services that published on the web have input/ output parameter of complex data type. In this research, we propose clustering approach that considers the simple type as well as complex data type in measuring the service similarity. We use hybrid term similarity method which we proposed in our previous work to measure the similarity. We capture the semantic pattern exist in complex data types and simple data types to improve the ontology learning method. Experimental results show our clustering approach which uses complex data types in measuring similarity works efficiently.

Construction of Web service recommendation systems for users has become an important issue in service computing area. Content-based service recommendation is one category of recommendation systems. The system recommends services based on functionality of the services. Current content-based approaches use syntactic or semantic methods to calculate the similarity. However, syntactic methods are insufficient in expressing semantic concepts and semantic content-based methods only consider basic semantic level. Further, the approaches do not consider the domain specific context in measuring the similarity. Thus, they have been failed to capture the semantic similarity of Web services under a certain domain and this is affected to the performance of the recommendation. In this paper, we propose domain specific context aware recommendation approach that uses support vector machine and domain data set from search engine in similarity calculation process. Experimental results show that our approach works efficiently.

Web service filtering is an efficient approach to address some big challenges in service computing, such as discovery, clustering and recommendation. The key operation of the filtering process is measuring the similarity of services. Several methods are used in current similarity calculation approaches such as string-based, corpus-based, knowledge-based and hybrid methods. These approaches do not consider domain-specific contexts in measuring similarity because they have failed to capture the semantic similarity of Web services in a given domain and this has affected their filtering performance. In this paper, we propose a context-aware similarity method that uses a support vector machine and a domain dataset from a context-specific search engine query. Our filtering approach uses a spherical associated keyword space algorithm that projects filtering results from a three-dimensional sphere to a two-dimensional (2D) spherical surface for 2D visualization. Experimental results show that our filtering approach works efficiently.

This paper addresses automatic service composition (ASC) as a means to create new value-added services dynamically and automatically from existing services in service-oriented architecture and cloud computing environments. Manually composing services for relatively static applications has been successful, but automatically composing services requires advances in the semantics of processes and an architectural framework that can capture all stages of an application's lifecycle. A framework for ASC involves four stages: planning an execution workflow, discovering services from a registry, selecting the best candidate services, and executing the selected services. This four-stage architecture is the most widely used to describe ASC, but it is still abstract and incomplete in terms of scalable goal composition, property transformation for seamless automatic composition, and integration architecture. We present a workflow orchestration to enable nested multilevel composition for achieving scalability. We add to the four-stage composition framework a transformation method for abstract composition properties. A general model for the composition architecture is described herein and a complete and detailed composition framework is introduced using our model. Our ASC architecture achieves improved seamlessness and scalability in the integrated framework. The ASC architecture is analyzed and evaluated to show its efficacy.

Organizing Web services into functionally similar clusters, is an efficient approach to discovering Web services efficiently. An important aspect of the clustering process is calculating the semantic similarity of Web services. Most current clustering approaches are based on similarity-distance measurement, including keyword, ontology and information-retrieval-based methods. Problems with these approaches include a shortage of high quality ontologies and a loss of semantic information. In addition, there has been little fine-grained improvement in existing approaches to service clustering. In this paper, we present a new approach to grouping Web services into functionally similar clusters by mining Web service documents and generating an ontology via hidden semantic patterns present within the complex terms used in service features to measure similarity. If calculating the similarity using the generated ontology fails, the similarity is calculated by using an information-retrieval-based term-similarity method that adopts term-similarity measuring techniques used by thesaurus and search engines. Another important aspect of high performance in clustering is identifying the most suitable cluster center. To improve the utility of clusters, we propose an approach to identifying the cluster center that combines service similarity with the term frequency-inverse document frequency values of service names. Experimental results show that our clustering approach performs better than existing approaches.

It is considered that Web services have had a tremendous impact on the Web as a potential silver bullet for supporting a distributed service-based economy on a global scale. However, despite the outstanding progress, their uptake on a Web scale has been significantly less than initially anticipated. The reasons are: first, the existing Web service frameworks such as gtraditionalh Web services, semantic Web services, and Web APIs have had a limited impact, second, isolated service islands without links to related services have hampered service discovery and composition. In this paper, we propose a methodology to drive innovation from isolated service islands into the global social service network to connect the islands. First, we propose Linked social service-specific principles based on Linked Data principles for publishing services on the open Web as linked social services using our new service model, and suggest a new platform for constructing a global social service network. Then, an approach is proposed to enable exploitation of a global social service network, providing Linked social service as a service. Finally, experimental results show that the Linked social service can solve the service discovery problem by enabling exploring service to service based on the global social service network.

In this paper, we propose a methodology to drive innovation from isolated service islands into the global social service network to connect the islands. First, we propose Linked social service-specific principles based on Linked Data principles for publishing services on the open web as linked social services using our new service model, and then an approach is proposed to enable exploitation of a global social service network, providing Linked social service as a service.

Automatic Service Composition (ASC) provides a new value-added service from existing services by user's request dynamically and automatically. User's requests consist of functional and nonfunctional requirements. During service composition, services that fulfill the functional requirements are located at the discovery stage. Abstract nonfunctional requirements should be identified mainly before the selection stage for service execution. Our research was motivated by the identification of abstract nonfunctional properties (NFPs) for a seamless ASC and proposes transformation from the abstract NFPs to intermediate-level NFPs based on the model of three levels of abstractness of NFPs. To solve the vagueness of the abstractness, we adapt approaches based not only on ontology but also on term similarity. The transformation between the intermediate and the concrete levels is carried out by a deterministic algorithm based on mapping of domain ontology. To evaluate the effectiveness of term similarity metrics for nonterminal terms, vector-based and large corpus-based approaches were investigated. The transformation performance based on precision over our test data set and ontology was evaluated.

New value-added services are created by automatic service composition (ASC) dynamically and automatically from existing services at the user's request. Complete ASC system requests solving very large and complex realistic problems require consistent architecture. Most studies of service composition are based on four stages (planning, discovery, selection, and execution) and their variations such as integration of the stages or divergence in a stage. However, previous studies have not considered the functional scalability of ASC involving nested dynamic services in which the ASC calls inner ASCs internally during composition. In practice, there are many situations where dynamic services and existing services are combined in ASC. We present a blueprint for a modified four-stage composition architecture to allow for scalability in managing the nested composition flow. A general model for the composition architecture is described and a complete and detailed composition framework is introduced using our model. Finally, we analyze the proposed architecture to show its efficacy.

Automatic Service Composition (ASC) supports creation of a new value-added composite service from the existing services with automatic manner. There are two approaches, machine-oriented and human-oriented, for the composer. In the machine-oriented composer, every composition step is managed by the composer mainly. However, there are many possibilities of interventions by users for better composition performance. In this paper, design of user interface for the machine-oriented ASC based on our new composition architecture is suggested. Possible interactions at the all the stages of service composition are analyzed on the new architecture. Ontologies for ASC UI, visual component, data, and workflow are designed. The Selector UI is demonstrated as an example of UI for ASC, and whole composition scenario is illustrated. The design paradigm of ASC UI presented in the paper can be applied to human-oriented composition too.

Semantic similarity measures play important roles in information retrieval and natural language processing. Several researches calculate semantic similarity between two words using web search engines as corpus or manually compiled corpus. In this paper, a method to find the word Ri between two words P and Q and extract a relation of the words with PLSI (Probabilistic Latent Semantic Indexing) is proposed. The results of the experiments show that using the PLSI with smaller latent class such is effective in getting Ri which is more related to P and Q, and using the PLSI with over 5 latent class is effective in getting veiled relation between P and Q.

The constraints revealed during a logical composition of services are often too abstract for automatic service composition. The abstract constraints have to be transformed to concrete attributes. This research investigates semi-automatic transformation of intermediate constraints to concrete constraints for automatic service composition. It considers simultaneously a stack of composition attributes for QoS, preferences, and logic constraints.

Solving general real-life problems requires a set of appropriate services to be composed via planning, scheduled, and then executed. Web service composition is the most difficult aspect and is our focus. In this paper, we describe a new framework for intelligent semantic Web services that supports the planning and scheduling aspects by a combined HTN planner and CSP. The framework covers all of the procedures needed to deal with a user's request, including domain analysis of the request, task flow decisions and CSP creation by the planner, and solving the CSP by a distributed CSP solver

In manufacture industry, better supply chain management (SCM) not only improves efficiency of business processes, but play important roles in a series of cost reductions. In a product manufacture, the initial design is more important from viewpoint of the cost reduction in the whole cycle. It becomes important how a designer collects the information on necessary parts for product design, and the system to search information efficiently at design stage is required. Information infrastructure to support design in SCM was developed on a semantic Web service environment that can provide interoperable service interfaces for integrated design attributes to agents and users. Design support agent (DSAgent), as a client for information infrastructure, helps the designer to find the desired product information. But this search process requires autonomy that lacks in DSAgent. As designers must change input values repeatedly to complete finding the desired product information, we suggest an autonomous DSAgent (ADSAgent) in this paper. To add autonomy, we used situation calculus, which is the schema for representing the dynamic changing world. The model to define the world was written in ConGolog statements, which is logic programming language based on situation calculus. This will facilitate user to find the desired product information and reduce user operations.