Service-Oriented Computing – ICSOC 2018 Workshops

Free-floating bike sharing (FFBS) is a new bike sharing mode when compared to the traditional station-based bike sharing (SBBS). It brings convenience for users since bikes can be picked up and returned anywhere but not the fixed stations. However, it also brings difficulty for managers because reallocation of free-floating bikes is totally different from any traditional ones. Using data-driven method, we define two types of nodes in this paper (i.e., easily and hardly accessed nodes), to represent different convenience levels of getting bikes from the FFBS. We collect bikes at hardly accessed nodes and reallocate them to the easily accessed nodes. Our objective is to move the needed bikes in the shortest distance and meanwhile to maximize the operation revenue. The problem is formulated as a multi-objective mixed integer programming model and an effective algorithm is designed to solve it. The test results can provide several constructive suggestions for reallocating the free-floating bikes.

Taking the airport de-icing resources optimization arrangements as the research object. When the airport de-icing resources are tense and large-scale delays occur at the airport, airport decision-makers need to consider both fairness and efficiency. At present, the FCFS method is widely used at home and abroad for de-icing scheduling, which has certain defects in fairness and efficiency. To improve the efficiency of airport de-icing and the fairness of resource allocation, through research and analysis of the theory and method of airport de-icing process scheduling problem, a multi-objective mathematical optimization model with the minimum number of strands as the efficiency goal and the minimum weighted dissatisfaction value as the fairness goal is established. An algorithm based on fully combination thinking is designed, and a multi-objective decision-making strategy is proposed for decision makers to choose a satisfactory solution as the airport de-icing resource allocation scheme. Based on the algorithm simulation of the model and comparative analysis, the results show that this algorithm is able to solve the problem of aircraft de-icing scheduling problem, which can improve the resource utilization efficiency better compared with the existing manual scheduling method and ensure the fairness of resource allocation to a certain extent.

Service modeling is an important step in designing service-oriented systems. There are multiple levels of design because service science includes both the business rationale and the IT implementation of the services. As business and IT perspectives differ, the modeling techniques are different, and often the respective modeling languages are disconnected or ad-hoc. We propose a new service-modeling approach for connecting the business modeling and the web service modeling by presenting these two perspectives in a single model. We present a multi-stage modeling process for capturing different perspectives and creating models iteratively by working with levels of abstraction from higher to lower. The model is then used as an input in order to generate a REST API specification in the OpenAPI format to feed the next stages of the service life-cycle.

Network virtualization can integrate the servers and computers from different locations in the large enterprises. Most of prior studies on the network virtualization execute on the cloud platform and they are not suit for the enterprise network. Therefore, the problem of the energy-aware and location-constrained virtual network embedding (EL-VNE) in the enterprise network is proposed and solved in this paper. Firstly, both the computing capability and bandwidth capability are unified by adopting the complex number theory and their corresponding capabilities of nodes, including physical and virtual nodes are determined. Then EL-VNE model is presented and proved to be a NP-complete problem, so as to make the virtual network embedding process only need node mapping without link mapping. Finally, a heuristic algorithm is presented to minimize the energy consumption on the condition of location constraint of nodes. The experiments show that the proposed EL-VNE can get less energy consumption compared with EAD, and simultaneously have better performance compared with GLC.

As an innovative Internet product grounded in Chinese particularistic culture, WeChat Red Envelops has received increasing attention from both academics and practitioners. A considerable number of studies have been conducted to investigate the WeChat Red Envelops phenomenon. However, the extant literature has been very scattered and little effort has been devoted to integrating the existing findings. Consequently, our understanding of the WeChat Red Envelops phenomenon is fragmented and limited. Against this background, this study conducted a systematic review of extant studies, summarized the current research status, identified the research gaps and accordingly provide future research directions. We believe that this study offers an important foundation for future research to conduct in-depth investigation and advance our understanding regarding WeChat Red Envelops.

In the medium or low frequency cases, the performance such as the gain or bandwidth of the fully differential operational amplifier is not good. In this paper, a fully differential telescopic two-stage operational amplifier with two-stage common-mode feedback is designed by using TSMC 250 nm technology. The first stage operational amplifier is telescopic structure, while the second stage operational amplifier is common source structure. The inner structure of common-mode feedback is continuous-time common-mode feedback circuit, and the outer structure is switched capacitor CMFB structure. It has the advantages of high gain and good linearity. It overcomes the shortcomings of common mode feedback circuit in limiting the output swing. It can stabilize the DC operating point and improve the output swing effectively. Cadence Spectre simulation results show that the common mode feedback structure keeps the common mode level at 1.25 V, the open loop gain is 67.7 dB, the phase margin is 45°, and the unit gain bandwidth is 150.7 MHz under the 2.5 V supply voltage.

The traditional convolutional neural networks are heavy with millions of parameters and the classification accuracy is not high. To address this issue, we propose a novel model called parallel concatenated convolutional neural network with cross-layer connections. The model mainly includes parallel processing and concatenate operation. In parallel processing, the diversity of features is increased by using different sizes of convolution kernels. The parallel outputs are integrated together by concatenate operation. Meanwhile, an improved cross-layer connection structure is also added to the model. At the experimental stage, the model was tested on the Caltech-256 and Food-101 datasets, the experiment results indicate that the constructed PCNet (without cross-layer connections) increases the recognition accuracy by 2.54% and 7.31% compared to AlexNet, and the proposed RPCNet (with cross-layer connections) is improved by 6.12% and 12.28% compared to AlexNet.

Abnormal user detection is one of the key issues in online social network security research. Attackers spread advertising and other malicious messages through stolen accounts, and malicious actions seriously threaten the information security of normal users with the credit system of social networks. For this reason, in the literature, there are a considerable amount of research work which detect abnormal accounts in social networks, however, these efforts ignore the problem of the seamless integration of machine learning with human behaviour-based analysis. This paper reviews the main achievements of abnormal account detection in online social networks in recent years from three aspects: behavioral characteristics, content-based, graph-based, and proposes a new social network abnormal user detection method based on system calls in computer’s kernel. Using enumeration sequence and hidden semi-Markov method, a hierarchical model of anomaly user detection in social networks is established.

In terms of image classification, in order to obtain higher classification accuracy, different levels of feature information need to be extracted from the image. Convolutional neural networks are increasingly applied to image classification. However, the traditional convolutional neural network has insufficient feature information extraction, poor classification accuracy, and easy over-fitting. This paper proposes Multi-branch aggregation network framework based on deep convolutional neural network that can be applied to image classification. Based on the traditional convolutional nerve, the network width and depth network are increased without increasing the parameters to optimize and improve the network to further enhance the feature expression ability of the network, Enriched the diversity of feature sampling, increased image classification accuracy and prevented overfitting. The framework and traditional frameworks and other frameworks were compared and analyzed through a series of comparative experiments in two standard databases, CIFAR-10 and CIFAR-100, and the validity of the framework was demonstrated.

Remote sensing images are blurred due to large and wide imaging, long shooting distance, fast scanning speed, interference from external light, etc. At the same time, because of that remote sensing images have the characteristics of diverse and dense shooting objects, deblurring remote sensing images is a major problem in remote sensing research. Therefore, we propose a remote sensing image deblurring algorithm, which based on WGAN. The algorithm is different from the traditional method in estimating the blur kernel of image. What’s more our method does not require an explicit estimation of the blur kernel, and it implements an end-to-end image deblurring process. We use a WGAN-based deblurring model. First, the training images are processed in pairs. Then, in order to increase the generalization ability, a image of 256 * 256 that is a sub-region cropped at the random position in the original image is chosen as the input image. Finally, to achieve a better deblurring effect, a content loss function and a perceptual loss function are added to the loss function to achieve the specific implementation. The remote sensing image deblurring model trained by the proposed method has achieved better results on the remote sensing image dataset. The experimental results show that the proposed algorithm have better performance than the traditional method in filtering out the blur of remote sensing images, which could optimize the overall visual effect subjectively and improve the peak signal-to-noise ratio of the image objectively.

In view of the traditional classification algorithm, the problem of high feature dimension and data sparseness often occurs when text classification of short texts. This paper proposes a text feature combining neural network language model word2vec and document topic model Latent Dirichlet Allocation (LDA). Represents a matrix model. The matrix model can not only effectively represent the semantic features of the words but also convey the context features and enhance the feature expression ability of the model. The feature matrix was input into the convolutional neural network (CNN) for convolution pooling, and text classification experiments were performed. The experimental results show that the proposed matrix model has better classification effect than the traditional text classification methods based on word2vec and CNN. In the text classification accuracy rate, recall rate and F1 three evaluation indicators increased by 8.4%, 8.9% and 8.6%.

Retinal image registration is the process of matching and superimposing two retinal images of the same patient. The traditional feature-based retinal image registration algorithm is computationally expensive during the matching process. This paper proposes a fast and efficient registration method based on the combination of bifurcation point and SURF algorithm. First, the eight-neighbor search algorithm is used to detect the bifurcation points of the reference image and the target image, and then the SURF feature is extracted in the rectangular template region centered on the bifurcation point. The Euclidean distance is used to perform rough matching on the extracted features, then RANSAC is used for fine matching, and finally the transformation model is estimated. Experiments show that this method can quickly and effectively achieve the registration of retinal images while reducing a large number of unnecessary searches and achieving a great registration result.

IoT is a combination of physical objects with virtual representations and services. Augmented reality provides an ideal interface to IoT applications by superimposing virtual information about smart objects and services on a user’s view of the real world. This allows a user to interact with the physical object as well as receiving additional context-aware information about the object e.g., size, speed and temperature, as well as information about nearby objects. However, users do not have to directly interact with the physical objects or sensors as augmented reality can be an effective method of providing additional information about IoT services in the environment, such as the QoS attributes of a service or the ratings that other users have provided during previous invocations. In this paper, we describe the use of augmented reality in IoT to provide contextual information to service users and providers and give a demonstration of an augmented reality application that we have developed and three projections of how users may interact with future context-aware applications. We also discuss some of the current research challenges and the future work that needs to be carried out to address these challenges.

With the widespread application of sensor and IoT technology in the field of water conservancy informatization, the traditional application systems based on Java EE or pure NoSQL databases for hydrological data processing and analysis have been difficult to meet the new requirements for processing and analyzing large-scale hydrological IoT stream data. How to select a suitable big data processing platform and how to implement application systems for hydrological IoT stream data requires in-depth theoretical foundations, more experimental comparisons, effective design paradigm and practical implementations. This paper summarizes the research status of big data in water conservancy domain, and then proposes a hydrological IoT data processing system based on Apache Flink. We use the sensor data obtained in Chuhe river as the experimental dataset, and take the common and daily operations for hydrological data as example. The experimental results show that the processing capability of the hydrological IoT data processing system is far superior to the traditional multi-tier architecture system based on Java EE or pure NoSQL databases, and it obviously becomes an appreciable solution for water conservancy informatization.

Although IoT service and service composition provide effective means to develop IoT applications, the dynamic and time-varying correlation among massive sensors rises up new challenges to the traditional model-based approaches, and the extra uncertainty and complexity of service composition become apparent. This paper proposes a data-driven service composition method based on our previous proactive data service model. We utilize real-time correlation analysis of sensor data to refine model-based service composition at runtime. The correlation among sensor data is usually asynchronous. In this paper, we adopt and improve a Dynamic Time Warping (DTW) algorithm to obtain one-way lag-correlation, and realize dynamic sensor data correlation through refining existing service composition. Based on the real sensor data set in a coal-fired power plant, a series of experiments demonstrate the effectiveness of our service composition method.

Service-oriented cloud-edge integration is a promising approach for the big IoT stream processing effectively in a distributed manner. Dynamic adaptation of cloud and edge service is of key importance to enable the seamless integration of cloud infrastructure and edge equipment. There exist several challenges, such as grasping the right moment and coping incomplete matching. Targeting at the challenges and based on our previous proactive data service model, the paper proposes a service adaptation approach, called as ALES, to enable the dynamic cloud-edge integration. The main contributions include: transforming the service adaptation problem into the improved maximal weight matching model in a dynamic bipartite graph. The M/M/c/∞ model in the queuing theory is modified to optimize the Kuhn-Munkres algorithm to minimize the average response time of the request of edge service. The effectiveness of the proposed approach is demonstrated by examining real cases of Chinese State Power Grid. Experimental results verify the effectiveness and efficiency of our approach.

With the rapid development of Internet of Things and smart services, massive intelligent devices are accessing the cloud data centers, which can cause serious network congestion and high latency issues. Recently, fog computing becomes a popular computing paradigm which can provide computing resources close to the end devices and solve various problems of existing cloud-only based systems. However, due to QoS (Quality of Service) constraints such as time and cost, and also the complexity of various resource types such as end devices, fog nodes and cloud servers, task scheduling in fog computing is still an open issue. To address such a problem, this paper presents a cost-effective scheduling strategy for multi-workflow with time constraints. Firstly, we define the models for workflow execution time and resource cost in fog computing. Afterwards, a novel PSO (Particle Swarm Optimization) based multi-workflow scheduling strategy is proposed where a fitness function is used to evaluate the workflow execution cost under given deadlines. A heart rate monitoring App is employed as a motivating example and comprehensive experimental results show that our proposed strategy can significantly reduce the execution cost of multiple workflows under given deadlines compared with other strategies.

Internet of Things (IoT) is a paradigm that has the capability to revolutionize on everyday life, in sectors of ranging from our homes, health, transport, industry, entertainment to interaction with government. The comfort and benefits that IoT is providing are undeniable, however, these may come with a huge risk of individual identity and data privacy. Several researches have been conducted to find a better way to eliminate privacy risks and minimize the effect on users’ privacy requirements. In this context, the proposed study consists of four segments, first we analyse the privacy problems evolving with the advancement in IoT paradigm. In the second segment, we present methodology review with analysis and classification of privacy solutions. Then we provide an in-depth analysis of preserving privacy during sensors data transmission. In the last segment, we depict the future trend for end-to-end privacy in IoT.

This paper presents a novel approach for urban road networks to estimate traffic speeds using vehicle trajectories captured by detectors on transportation cells. By scanning and analyzing dynamic traffic streams of passing-vehicles, we calculate the real-time traffic speed of road segment separated by adjacent detectors, which are further synthesized to present the traffic speed of whole road. Compared to driving routes data with limited coverage or floating GPS data with occasional missing that are both frequently utilized for traditional road speed estimation, our approach utilizes the full coverage detector data and is proved to have more accurate and reliable results in its application for two large cities of China. An analysis and visualization system was hence developed, whose successful operation in several transportation departments indicated the efficiency of our approach. It helps to guide travelers the optimal driving routes, which greatly relieves the huge traffic stress of city road.

With the development of highway toll system and sensor network, massive highway toll data has been accumulated nowadays. The imperfection of raw data, such as incomplete, repetitive and abnormal data, seriously affects the efficiency of data mining modeling. Traditional cleaning methods on massive spatio-temporal data are inefficient, because the business rules are difficult to depict in various domains. On the highway toll data of Henan Province, we propose a data cleaning service through business rules. This service can efficiently clean the raw toll data with spatio-temporal attributes, including the data calibration of erroneous data and invalid data, the repair of erroneous data, and the filtering of duplicate data. Implemented through Hadoop MapReduce on toll data in highway domain, our service shows its efficiency, accuracy and scalability in extensive experiments.

This paper proposes to use machine learning techniques with ultrasonic sensors to predict the behavior and status of a person when they live solely inside their house. The proposed system is tested on a single room. A grid of ultrasonic sensors is placed in the ceiling of a room to monitor the position and the status of a person (standing, sitting, lying down). The sensors readings are wirelessly communicated through a microcontroller to a cloud. An intelligent system will read the sensors values from the cloud and analyses them using machine learning algorithms to predict the person behavior and status and decide whether it is a normal situation or abnormal. If an abnormal situation is concluded, then an alert with be risen on a dashboard, where a care giver can take an immediate action. The proposed system managed to give results with accuracy exceeding 90%. Results out of this project will help people with supported needed, for example elderly people, to live their life as independent as possible, without too much interference from the caregivers. This will also free the care givers and allows them to monitors more units at the same time.

In this paper, we proposed an efficient R-tree construction method by bulk loading over spatial-temporal data stream. The core idea is to partition spatial-temporal data stream into time windows and construct an R-tree for each time window. In each time window, we parallelized space partitioning and data stream reception during R-tree construction; and then we adopted sorting-based bulk loading scheme to optimize R-tree construction, which avoided unnecessary synchronization overhead and accelerated the R-tree construction. In addition, to reduce the sorting cost of R-tree bulk loading, sampling-based space partitioning scheme was introduced. Theoretical analysis and experiments demonstrated the effectiveness of our proposed method.

Cloud environment provides high performance computing services to process massive data for data-intensive workflows. Due to the different functional requirements, tasks in a workflow might be allocated to multiple cloud servers. The massive data among these tasks have to be transferred and this greatly increases the execution cost. To decrease the transferred data size during the workflow execution, this paper proposes a dynamic task allocation method based on the data dependencies. The workflow with data dependencies and typical control logic, i.e., sequential, parallel, and exclusive choice, is described based on process algebra. The data size relevant to a data dependency can be obtained only after the task is executed. Each task is allocated to a certain server according to relevant data size and maximal data paths. A case study is presented to illustrate the feasibility and effect of the proposed method and the related work is discussed based on the case study.

Data replication is one of the significant sub-areas of data management in cloud based workflows. Data-intensive workflow applications can gain great benefits from cloud environments and usually need data management strategies to manage large amounts of data. At the same time, multi-cloud environments become more and more popular. We propose a cost-effective and threshold-based data replication strategy with the consideration of both data dependency and data access times for data-intensive workflows in the multi-cloud environment. Finally, the simulation results show that our approach can greatly reduce total cost of data-intensive workflow applications by considering both of data dependency and data access times in multi-cloud environments.

The number of interconnected smart devices has already rapidly increased, and the Internet of Things (IoT) has presented tremendous potential in various domains such as smart cities, healthcare and industrial automation. To integrate the IoT applications to Web to utilise the advantages of Internet infrastructures, the Constrained Application Protocol (CoAP) is proposed as one of the standardised protocols for IoT applications. However, the REST architecture style, which is the foundation of Web, was not designed for IoT applications and thus cannot satisfy all the requirements of IoT applications. To efficiently monitor the IoT resources asynchronously, the IETF (Internet Engineering Task Force) extended the CoAP with Resource Observe mechanism. However, the Resource Observe mechanism benefits sensors rather than actuators. For the actuator resources, the CoAP cannot support the context-adaptation, and therefore it cannot always correctly estimate system states and handle complex physical behaviours. In this paper, we extend the CoAP with a context-adaptation mechanism to enrich the system states estimation and other operations in the protocol level for physical behaviour modelling and implementation. The extended mechanism is implemented in the Californium (CF) framework.

In scenarios involving text classification where the number of classes is large (in multiples of 10000 s) and training samples for each class are few and often verbose, nearest neighbor methods are effective but very slow in computing a similarity score with training samples of every class. On the other hand, machine learning models are fast at runtime but training them adequately is not feasible using few available training samples per class. In this paper, we propose a hybrid approach that cascades (1) a fast but less-accurate recurrent neural network (RNN) model and (2) a slow but more-accurate nearest-neighbor model using bag of syntactic features.Using the cascaded approach, our experiments, performed on data set from IT support services where customer complaint text needs to be classified to return top-N possible error codes, show that the query-time of the slow system is reduced to $$1/6^{th}$$ while its accuracy is being improved. Our approach outperforms an LSH-based baseline for query-time reduction. We also derive a lower bound on the accuracy of the cascaded model in terms of the accuracies of the individual models. In any two-stage approach, choosing the right number of candidates to pass on to the second stage is crucial. We prove a result that aids in choosing this cutoff number for the cascaded system.

A successful taxi route recommendation system is helpful to achieve a win-win situation for both increasing drivers’ income and improving passengers’ satisfaction. The critical problem in this system is how to find the optimal routes under the highly time-varying and complex traffic environment. By investigating the main factors and comparing various route recommendation methods, in this paper, we handle the taxi route recommendation issue from a new perspective. The relationships between the cruising taxis and passengers are regarded as attraction or repulsion between electric charges. Then based on urban traffic charge heat map, we propose a simple yet effective taxi route recommendation method named TaxiC. TaxiC considers four key factors: the number of passengers, travel distance, traffic conditions, vacant competition, and then recommends driving direction in real time for drivers to help them find the next passengers more efficiently and reduce the cruising time. The experimental results on a real-world data set extracted from 5398 taxis in Xiamen city demonstrate the effectiveness of the proposed method.

Poor quality of process event logs prevents high quality business process analysis and improvement. Process event logs quality decreases because of missing attribute values or after incorrect or irrelevant attribute values are identified and removed. Reconstructing a correct value for these missing attributes is likely to increase the quality of event log-based process analyses. Traditional statistical reconstruction methods work poorly with event logs, because of the complex interrelations among attributes, events and cases. Machine learning approaches appear more suitable in this context, since they can learn complex models of event logs through training. This paper proposes a method for reconstructing missing attribute values in event logs based on the use of autoencoders. Autoencoders are a class of feed-forward neural networks that reconstruct their own input after having learnt a model of its latent distribution. They suit problems of unsupervised learning, such as the one considered in this paper. When reconstructing missing attribute values in an event log, in fact, one cannot assume that a training set with true labels is available for model training. The proposed method is evaluated on two real event logs against baseline methods commonly used in the literature for imputing missing values in large datasets.

As the number of users and locations has increased dramatically in location-based social networks, it becomes a big challenge to recommend point-of-interests (POIs) meeting users’ preference. In traditional recommendation tasks, personalized recommendations performs well, however, these methods also have many disadvantages such as the long-tailed problem and the strong assumption. Further, in general scenarios, a group of users (e.g., colleagues, friends, and family members) often visit a specific location to enjoy time together (e.g., meal and shopping). Thus, it is more meaningful to recommend locations to the group than to individuals. However, the existing group recommendation approaches also have some limitations that hardly capture the preferences of a group of users effectively. To make full use of the users’ preferences and improve the effectiveness of group recommendation, in this paper, we propose a multi-layer ensemble framework which has a two-step fusion process. For the first step, we employ several personalized recommendation methods to generate the recommendations for individuals, and the recommendation list is obtained using the proposed fusion approach based on the supervised learning. For the second step, we utilize several ranking aggregation algorithms to fuse the recommendations list of individuals in the group and propose an unsupervised learning based ranking algorithm (URank) to further fuse the results of ranking aggregations to obtain the final group recommendation list. The experiments are conducted on a real-world dataset, and the results demonstrate the effectiveness of our proposed general framework.

The evaluation of business process predictive monitoring models usually focuses on accuracy of predictions. While accuracy aggregates performance across a set of process cases, in many practical scenarios decision makers are interested in the reliability of an individual prediction, that is, an indication of how likely is a given prediction to be eventually correct. This paper proposes a first definition of business process prediction reliability and shows, through the experimental evaluation, that metrics that include features defining the variability of a process case often give a better prediction reliability indication than metrics that include the probability estimation computed by the machine learning model used to make predictions alone.

In the Internet of Things (IoT), billions of physical devices connecting over the Internet provide a near real-time state of the world in a service-oriented way. The demand-driven service-provision paradigm may need a negotiation process to tailor the service properties before creating the service level agreement (SLA). Existing negotiation techniques are focused on the cloud computing, however, SLA negotiation is rarely discussed in the IoT environment. Thus, we extended a commonly-used web service negotiation framework based on characteristics of the IoT, integrated with a game theory-based negotiation strategy, and evaluated its performance under a simulation platform. Based on the result, we identified the research questions and outlined future directions.

As the number of IoT devices with limited resources and the corresponding observed data grow exponentially, the method of offloading all tasks to a remote data center becomes expensive, even inefficient. How to optimize the energy consumption of application requests from IoT devices satisfying the deadline constraint is also a challenge. Fog computing is closer to users, featuring the lower service delay but less resource than the remote cloud. Fog does not mean to replace cloud. They are complementary to each other, and cooperation between them is worth studying. The main points of this paper are: (1) Proposing a general IoT-fog-cloud computing architecture that fully exploits the advantages of fog and cloud. (2) Formulating the energy efficient computation offloading and dynamic resource scheduling (eoDS) problem, then proposing an eoDS algorithm to solve the problem, reducing the energy consumption and completion time of application requests (3) Compared with cloud nodes, the mobility of fog nodes is higher. For this, we propose the fog functional areas reconstruction method to adaptively deal with the changing environment, improving the resource utilization of fog.

The resources of mobile application in the app stores contains a vast amount of code and knowledge, which is of great significance to intelligent program search technology, but how to organize and utilize these multi-source heterogeneous data efficiently and integrate semantic information is still a key problem. In this paper, WordNet based AppNet, a multi-layer heterogeneous complex app network model was proposed, it completely describes the hierarchical structure between app-related tags, attributes, and code and in which aims to explore its application in the intelligent program search. Firstly, we expound the construction mechanism of AppNet and describe how does it realize mapping with WordNet, and then two simple real application scenarios were conducted based on AppNet in which to verify its validity and feasibility. We believe that the proposed AppNet model will provide researchers with more efficient ideas in the field of intelligent software development and search.

Cognitive service is an emerging service paradigm in service-oriented computing. It can comprehend data in the same way as the human. Cognitive services require sufficient information to understand service scenarios. Actually, to achieve a goal sometimes requires multiple services with order dependencies and prerequisites to work collaboratively. When an exception event occurs during the service group working, the conventional approach is to restart or stop the service based on the exception type. If the environment information is changed much fast and retrieved unpractically, the exception event can cause the delayed response of the service group. If the goal of the service group is time-aware and service result is preferred, the regular policy is hard to match the requirement. In this paper, we address the problem of delayed response caused by exception events raised from the distributing cognitive service group. A novel architecture is proposed to ensure the overall consistency and real-time reaction of distributing cognitive service group.

Nowadays, crossover and convergence between services has become a new phenomenon in service market, especially in China. This work aims at analyzing the inner and external formation mechanism of crossover service, which have great significance to the sustainable development of this new ecosystem. Previously, a traditional concept based knowledge graph extracted from crossover service News was constructed for crossover service event analysis. The crossover modes and the statistics of appearances were found in that experiment. Due to the limitation of motivation analysis of the inference method based on static concept structure, the Event Evolutionary Graph (EEG) is introduced to improve the event analysis, and moreover to achieve the event prediction in this undergoing experiment. Generic EEG can represent the event evolution rules and patterns, but due to the characteristics of crossover services, this experiment will adapt event modeling method and its training method. This work in progress methodology would explain the inner and external formation reason of crossover service, and furthermore achieve the assisted decision-making for potential crossover organizations based on event prediction.

Manually reasoning about candidate refactorings to alleviate bottlenecks in service-oriented systems is hard, even when using high-level architecture/deployment models. Nevertheless, it is common practice in industry. Elmo is a decision support tool that helps service-oriented architects and deployment engineers to analyze and refactor architectural and deployment bottlenecks in service-oriented systems.

The IT service management is transforming or evolving with artificial intelligence. A data-driven and knowledge-based approach potentiates the IT service management optimization and automation with the goal of delivering better business outcomes. In this demo, we show our framework, cognitive event automation (CEA), that applies artificial intelligence to the automated resolution of incident tickets, and the methodology and technologies for creating knowledge by analyzing tickets, eliminating those that do not require action as well as auto-resolving those that do. The case study shows CEA can help the IT service management system to deliver better business outcomes.

Migration of legacy applications to Cloud has been growing steadily over the past years, driven by the promise of greater flexibility, scalability, and lower management costs. However, the complexity of the migration tasks and activities makes transformation of the current service and application architectures a long and difficult process that involves months of migration planning and execution. In this paper, we present a service application BluePlan and its implementation, which employs an artificial intelligence (AI) planner that optimizes the end-to-end migration planning with constraints, and creates migration plans for execution. The AI planner service serves to expedite and simplify the migration planning process by defining the clients’ constraints and resources in a simplified format that abstracts the user’s need to hardcode domains and problems. This capability is exposed as a service and evaluated for migration plans for over 500 hundred clients with varying independent memory, cpu and time constraints in the span of a few minutes, thereby enabling migration project manager and migration architects to reason about potential migration plans, and replan as needed.

As software architecture design is evolving to microservice paradigms, RESTful APIs become the building blocks of applications. In such a scenario, a growing market of APIs is proliferating and developers face the challenges to take advantage of this reality. For example, third-party APIs typically define different usage limitations depending on the purchased Service Level Agreement (SLA) and, consequently, performing a manual analysis of external APIs and their impact in a microservice architecture is a complex and tedious task. In this demonstration paper, we present ELeCTRA, a tool to automate the analysis of induced usage limitations in an API, derived from its usage of external APIs. This tool takes the structural, conversational and SLA specifications of the API, generates a visual dependency graph and translates the problem into a constraint satisfaction optimization problem (CSOP) to obtain the optimal usage limitations.

With the development of artificial intelligence (AI) industry, the research field is more and more closely related to AI industry. It is increasingly important to analyze the development trend of AI, predict and grasp the research direction, and provide industrial transformation in time. There are many kinds of AI classification, and the scale of literature data is also very large. In this paper we propose a visual analysis system based on micro-service architecture, which is an efficient, robust and easily extended and easy-to-use visual platform for AI development situation analysis. Our paper describes the usage of the system in detail.

It is challenging to analyze the control flow relations of tasks in a complex business process. To solve this problem, quite a few process abstraction and reduction techniques have been proposed. However, existing approaches are lacking of the support for network structures in control flows. In this demonstration, we present a graphical Web application, TReAT (Tasks Relation Analyzing Tool). With TReAT, users can model their business processes and analyze relations between tasks in a process. Most common control flow structures including the network structure are supported by TReAT.

Analyzing data-driven and knowledge intensive business processes is a key endeavor for today’s enterprises. Recently, the Internet of Things (IoT) has been widely adopted for the implementation and integration of data-driven business processes within and across enterprises. For example, in law enforcement agencies, various IoT devices such as CCTVs, police cars and drones are augmented with Internet-enabled computing devices to sense the real world. This in turn, has the potential to change the nature of data-driven and knowledge intensive processes, such as criminal investigation, in policing. In this paper, we present a framework and a set of techniques to assist knowledge workers (e.g., a criminal investigator) in knowledge intensive processes (e.g., criminal investigation) to benefit from IoT-enabled processes, collect large amounts of evidences and dig for the facts in an easy way. We focus on a motivating scenario in policing, where a criminal investigator will be augmented by smart devices to collect data and to identify devices around the investigation location and communicate with them to understand and analyze evidences. We present iCOP, IoT-enabled COP assistant system, to enable IoT in policing and to accelerate the investigation process.

In the era of big data, the quality of services any organization provides largely depends on the quality of their data-driven processes. In this context, the goal of process data science, is to enable innovative forms of information processing that enable enhanced insight and decision making. For example, consider the data-driven and knowledge-intensive processes in Australian government’s office of the e-Safety commissioner, where the goal is to empowering all citizens to have safer, more positive experiences online. An example process, is to analyze the large amount of data generated every second on social networks to understand patterns of suicidal thoughts, online bullying and criminal/exterimist behaviour. Current processes leverage machine learning systems, e.g., to perform automatic mental-health-disorders detection from social networks. This approach is challenging for knowledge workers (end-user analysts) who have little knowledge of computer science to use machine learning solutions in their data-driven processes. In this paper, we present a novel platform, namely iSheets, that makes it easy for knowledge workers of all skill levels to use machine learning technology, the way people use spreadsheet. We present and develop a Machine Learning (ML) as a service framework and a spreadsheet-based ML development platform to enable knowledge workers in data-driven processes engage with ML tasks and uncover hidden insights through learning in an easy way.

Continuous integration (CI) is a key practice where developers integrate frequently via a shared repository to enable automated build, test, and release of software systems. While enabling CI in a centralised development environment has been a common practice, no much work has been done to effectively support CI of decentralised service-oriented systems where centralised repositories are unavailable. This paper presents SORCER, a decentralised interface-based continuous integration platform that makes it easy for developers to perform integrated build and test of service-oriented systems whose service constituents are owned and managed by different organisations to only expose their interfaces without access to their source codes.

In this demonstration, we design and implement a prototype of proof for causal graph building, anomaly detection and root cause analysis of microservice systems. The system comprises two core functionalities: (i) monitoring of systems and services; (ii) Application anomaly detection and root cause analysis. In the first part, the key metrics for the health of a system and an application, are collected by backend and plotted with dynamic charts in the frontend, which can help operators spot the overall system status. In the second part, the system can automatically build a causal graph of the microservice applications, indicating the dependencies between different modules, without instrumenting any source code. When an anomaly of a service instance is detected, it will be highlighted in the graph. A root cause inference function is also applied to analyze the root cause and returns a ranked list of root cause candidates to operators.

REST has become the architectural style of choice for APIs, where clients need to instantiate a potentially lengthy sequence of requests to the server in order to achieve their goal, effectively leading to a RESTful conversation between clients and servers. Mining the logs of such RESTful conversations can facilitate knowledge sharing among API designers regarding design best practices as well as API usage and optimization. In this demo paper, we present the RESTalk Miner, which takes logs from RESTful services as an input and uses RESTalk, a domain specific language, to visualize them. It provides interactive coloring to facilitate graph reading, as well as statistics to compare the relative frequency of conversations performed by different clients. Furthermore, it supports searching for predefined patterns as well as pattern discovery.

Due to lack of standardization and automation, large-scale Service Level Agreement (SLA) management remains challenging for IT service providers. For instance, flexible re-use of SLA definitions across different client engagements is often poorly supported by current SLA management frameworks. In this demonstration we present the ysla Engine, a new SLA management framework implementing the YAML-based ysla language for modeling SLAs. ysla provides novel semantic constructs for adaptable SLA templates that formally separate metrics definitions from associated customer-specific classification/categorization taxonomies for monitored subjects. In our demonstration we model the common, but intricate industry use case of SLAs for incident management and demonstrate how ysla-based SLA templates and SLAs can foster cross-client SLA re-usability.

The road to fully and secure autonomous cars is still long and exceedingly complicated. For instance, like smart city and virtual reality, self-driving cars need the infrastructure and data networks to catch up before it become common, safe and widely used. 5G telco network is considered as the required key concept that could enable autonomous cars operation. It would enable edge analytics and intelligence capabilities that are still missing in the current autonomous cars. This paper proposes one more slice of the next-generation self-driving automobile. It introduces a prototype that implements an autonomous car traveling into smart city. The car relies on emerging computing models such as Multi-access Edge Computing to perform part of the computation at the edges of 5G, in the surroundings of the car. The ultimate goal is to reduce latency and allow the car to make rapid decisions during the trip.

Juno is a chat-based service that interacts with user through natural language, in order to understand, assist and execute the user’s service request with a IT Service Management (ISM) system.