To address the problems of large volume of science and technology information, low information value density, and matrix sparsity of recommendation algorithms, we propose STIR-KG, a science and technology information recommendation method integrating knowledge graph, and build a science and technology information recommendation service. The main contributions are: (1) Establishing a new material knowledge graph, which has been open-sourced in GitHub (2) Combining collaborative filtering methods with knowledge graphs to solve the cold-start and matrix sparsity problems. (3) Propose the representation learning method TransAR, which enhances the representation capability compared with traditional methods, and uses the Mahalanobis distance metric score function to reduce the influence of irrelevant dimensions on the similarity calculation. (4) Based on the STIR-KG method, we use the streaming computing framework Flink to build a recommendation service for scientific and technical information, which captures user interest migration in real time and makes the recommendation results more time-efficient. And according to the experimental verification, STIR-KG has significantly improved the accuracy and recall rate compared with other algorithms.

Motivated by the requirement of system-level anomaly prediction in the running of industrial processes, this paper proposes a new algorithm to mine temporal dependencies among services, by discovering frequent occurrence patterns among service outputted events. With temporal dependencies, the paper also explores a new type of graph-based service linking approach. These approaches are delivered to prediction of system-level anomalies in some real scenarios.

Information service is always a hot topic especially when the Web is accessible anywhere. In university, lecture information is very important for students and teachers who want to take part in academic meetings. Therefore, lecture news extraction is an important and imperative task. Many open information extraction methods have been proposed, but due to the high heterogeneity of websites, this task is still a challenge. In this paper, we propose a method based on fusing multiple features to locate lecture news on the university website. These features include the linked relationship between parent webpage and child webpages, the visual similarity, and the semantics of webpages. Additionally, this paper provides an information service based on a main content extraction algorithm for extracting the lecture information. Stable and invariant features enable the proposed method to adapt to various kinds of campus websites. The experiments conducted on 50 websites show the effectiveness and efficiency of the provided service.

The rise of the Internet of Things (IoT) devices and the streaming platform has tremendously increased the data in motion or streaming data. It incorporates a wide variety of data, for example, social media posts, online gamers in-game activities, mobile or web application logs, online e-commerce transactions, financial trading, or geospatial services. Accurate and efficient forecasting based on real-time data is a critical part of the operation in areas like energy & utility consumption, healthcare, industrial production, supply chain, weather forecasting, financial trading, agriculture, etc. Statistical time series forecasting methods like Autoregression (AR), Autoregressive integrated moving average (ARIMA), and Vector Autoregression (VAR), face the challenge of concept drift in the streaming data, i.e., the properties of the stream may change over time. Another challenge is the efficiency of the system to update the Machine Learning (ML) models which are based on these algorithms to tackle the concept drift. In this paper, we propose a novel framework to tackle both of these challenges. The challenge of adaptability is addressed by applying the Lambda architecture to forecast future state based on three approaches simultaneously: batch (historic) data-based prediction, streaming (real-time) data-based prediction, and hybrid prediction by combining the first two. To address the challenge of efficiency, we implement a distributed VAR algorithm on top of the Apache Spark big data platform. To evaluate our framework, we conducted experiments on streaming time series forecasting with four types of data sets of experiments: data without drift (no drift), data with gradual drift, data with abrupt drift and data with mixed drift. The experiments show the differences of our three forecasting approaches in terms of accuracy and adaptability.

Many anomaly detection algorithms have been provided as services for more convenient and efficient utilization in IoT era. Due to concept drift existed in dynamic IoT stream data, it is a changeling task to apply proper anomaly detection services at run time. For effective on-line anomalous data discovery, this paper proposes a service selection framework to dynamically select and configure anomaly detection services. A fast classification model based on XGBoost is trained to identify the pattern of various stream data, so that suitable service can be selected and configured according to the pattern of stream data. Extensive experiments on real and synthetic data sets show that our framework can select suitable service for different scenarios, and the accuracy of the chosen services outperforms state-of-the-art methods.

Information service is always a hot topic especially when web is accessible anywhere. In university, lecture information is very import for students and teachers who want to take part in academic meetings. Therefore, lecture news extraction is an important and imperative task. Although many open information extraction methods have been proposed, but due to the highly heterogeneity of website, this task is still a challenge. In this manuscript, we propose a method based on fusing multiple features to locate lecture news in university web site. These features including the organization structure of lecture news catalog webpage, the visual similarity and the semantic of webpage. Additionally, this paper provide an information service based on a main content extraction algorithm for extracting lecture information. The stable and invariant features enable the propose method could adaptive to many kinds of campus website. The experiments conducted on 50 websites show the effectiveness and efficiencies of provided service.

Most of Traditional Chinese Medicine (TCM) data and ancient records exist in the form of books. The unstructured medical information is the foundation for building TCM knowledge service. The existing methods are not accurate enough to solve TCM named entity recognition and require a lot of manual labeling data. This paper proposes a semi-supervised embedded Semi-BERT-BiLSTM-CRF model. Based on the book “Diagnosis of Traditional Chinese Medicine in Traditional Chinese Medicine”, we select the physical features from the cleaned-up text information according to the characteristics of Chinese medicine classics, and then use a small amount of labeled data to train the BERT-BiLSTM-CRF model. The obtained model is used to predict unlabeled data and obtain pseudo-label data. The pseudo-label and labeled data are used as a training set for model training. Experiments show that TCM entity recognition accuracy of this method reaches 81.24%, which effectively improves the TCM entity recognition accuracy and reduces the manual labeling work. The results of this research can be applied to scenarios such as auxiliary diagnosis of TCM and expert system after subsequent improvement and transformation.

Granger causality and its learning algorithms have been widely used in many disciplines to study cause-effect relationship among time series variables. In this paper, we address computing challenges of state-of-art Granger causality learning algorithms, specially when facing increasing dimensionality of available datasets. We study how to leverage gradient boosting meta machine learning techniques to achieve accurate causality discovery and big data parallel techniques for efficient causality discovery from large temporal datasets. We propose two main algorithms for gradient boosting based causality learning, and parallel gradient boosting based causality learning. Our experiments show our proposed algorithms can achieve efficient learning in distributed environments with good learning accuracy.

The complex interventions among sensor streams bring new challenges for IoT applications to derive meaningful information from large amounts of sensor streams. This paper aims to provide a data-driven service creation method for effectively capturing events based on our previous service abstraction – proactive data service. For improving the effectiveness of proactive data service, we consider the potential correlations among sensor streams besides user's pre-definitions when creating service. Based on the assumption that events frequently co-occurred in history have high probability to co-occur again, we regard frequent event sets as one kind of correlations among sensor streams, and propose an algorithm called FP-MFIM to efficiently find the maximum frequent event sets co-occurred in multiple sensor streams. For providing more effective information, we create PD-services with frequent co-occurred event types besides user-defined event types. This paper reports the tryout use of the method in China power grid for power quality event detection and location. Through a series of experiments based on real sensor data from power grid, we verified the efficiency of FP-MFIM algorithm and the effectiveness of our PD-services in real-world scenario.

In an IoT (Internet of Things) environment, event correlations may be dynamically interwoven because events usually span over many interrelated sensors. Our previous works used frequent sequence to measure the event correlations and mined them from a statistical perspective. We also proposed a service hyperlink model to encapsulate the event correlations. With the service hyperlink model, we made a preliminary attempt to reuse valuable event correlations to facilitate IoT applications. To consummate our previous method, this paper refines it in reusing the correlations, and focuses on which event would most probably appear after a previous event has occurred. To effectively mine the event correlations, we extend the traditional motif mining algorithms by introducing time constraint. Moreover, we connect services by service hyperlinks (i.e., abstraction of event correlations) to form a directed graph. An event can be routed on this graph. We have applied our approach in anomaly warning in a coal power plant and made extensive experiments to verify the effectiveness of the approach.

Edge computing may improve the processing quality of big IoT stream data and reduce network operational cost by moving computation onto the edge. However, there are two challenges in integrating cloud and edge computing for big stream data. Firstly, edge equipment usually has very limited computing power as well as storage ability, and apparently cannot support all the processing of big and real-time stream data. A flexible division of such services between edge and cloud is needed. Secondly, edge-end collaboration continuously changes due to some intrinsic interaction of data stream. In this paper, we propose a service-based approach to seamlessly integrating cloud and edge equipment. Based on our service model, we split a cloud service into two parts running on cloud and edge respectively. Also, we propose a dynamic service scheduling mechanism based on the improved bipartite graphs. We can deploy a cloud service to the edge at the right time when a key node emerges. The effectiveness of the proposed approach is demonstrated by examining real cases of China's State Power Grid. Experimental results verify the effectiveness and efficiency of our approach.

Predictive industrial maintenance promotes proactive scheduling of maintenance to minimize unexpected device faults. A fault is not always isolated and may be formed by a propagation of trivial anomalies, which are regarded as service events herein. In this paper, we firstly propose an algorithm for generating service event correlation. Such correlations can show us lots of clues to the anomaly/fault propagation. The correlations are encapsulated into service hyperlinks as our previous works did, and thus we depict the anomaly/fault propagation as service event routing among services via the refined service hyperlinks. Our scenario illustrates that a trivial anomaly may propagate into different faults under different service event correlations. It indicates that the destination of a service event is often uncertain. Therefore, this paper further proposes a heuristic approach to handle the uncertainty problem. Extensive experiments have been made to verify the effectiveness of the approach.

A surge in sensor data volume has exposed the shortcomings of cloud computing, particularly the limitation of network transmission capability and centralized computing resources. The dynamic intervention among sensor streams also brings challenges for IoT applications to derive meaningful information from multiple sensor streams. To handle these issues, this paper proposes a service‐based method with fog computing paradigm based on our previous service abstraction, which can capture meaningful events from multiple sensor streams. In our service abstraction, we utilize correlation analysis method to capture events as variations of correlation among sensor streams. Facing inconsistent frequency and shift of correlation, we propose a Dynamic Time Warping‐ (DTW‐) based algorithm to obtain sensor streams’ lag‐correlation. For adaptively aggregating related events from different services, we also propose an event routing algorithm to assist the composition of cascaded events through service collaboration. This paper reports the tryout use of our method in Chinese power grid for detecting abnormal situations of power quality. Through a series of experiments based on real sensor data in power grid, we verified that our method can reduce the network transmission and computing resource with high accuracy.

To detect anomalies according to system log is a hot topic recently. For the harmonic monitoring system of the power grid, the common practice of anomaly detection is to conduct machine learning. The learning model is trained with the historical anomaly data, and used for online detection. The premise of this method is to predefine a set of indicators as the input features of the machine learning model. However, existing methods rely mainly on business experience to extract such indicators, which limits the scope of the indicators used for data analysis, but also limits the accuracy of power quality perturbation analysis. In this paper, we propose an algorithm for power quality disturbance detection which investigates the correlation among the harmonic monitoring indicators, and extract the frequently concurrent abnormal indicators as the features to locate power quality disturbance detection. With the verification of the historical disturbance records, we prove that our algorithm can effectively detect the power quality disturbing events.

Abnormality Detection in power plant is a typical IoT application which aims to identify anomalies in these routinely collected monitoring sensor data; intend to help detect possible faults in the equipment. However, on the development of abnormality detection, we find that there are three challenges. The first one is the lack of cooperation between sensors. It means that the physical sensors cannot share and interact with each other. Secondly, the rapid increase in volume of sensor data and dynamic situation of production result in challenges to predefine all possible associations between sensors. Thirdly, it is difficult to build IoT application for developers who have little or no professional knowledge about production process. In this paper, we proposed a proactive data service model to encapsulate stream sensor data into services. We spread events among the proactive data services. By analysis of event correlations, we have realized service hyperlinks which help to offer the proactive real-time interaction with services. Real application and experiments verified that our proactive data service based method is more effective compare with traditional rule-based methods to detect abnormalities in power plant.

IoT service and service composition provide an effective means to develop IoT applications based on correlating multiple sensor data. The change of specific sensor data can cause others' changes under uncertain situations. It makes difficult for defining service composition plan in advance to build IoT application. This paper proposes a data-driven service composition method based on our previous proactive data service model. We regard service events frequently happen together with given service event as its situation, and the service events happen next as reacted actions under the situation. We analyze two kinds of correlation among service events via an improved FP-tree algorithm, and realize the service composition at runtime based on the real-time service events. Based on the real sensor data set in a coal-fired power plant, a series of experiments demonstrate that our method can effectively detect new service events based on current service events.

In an IoT environment, process analysis becomes more difficult as a process usually spans over a set of autonomous and distributed sensors. This paper consummates our previous service hyperlink model, to encapsulate dependencies among events generated from services. To effectively discover service hyperlinks, we transform the service hyperlink discovery problem into a frequent sequence mining problem. Existing frequent sequence mining algorithms cannot be directly used because they do not take the temporal constraints in event dependencies into consideration. Based on the dataset from a real power plant as well as several synthetic datasets, we do lots of experiments to verify the effectiveness and efficiency of our algorithm.

With the increasing abundance of traffic data from sensors and devices, the integration and analysis of such streaming data are gaining importance in many application scenarios. This paper proposes a service-based approach for integrating and analyzing the traffic sensor data to support green commute in China by automatically discovering carpooling companions within a community. Our focuses are on modeling and design of the carpooling discovery services, algorithms for implementing the services, and performance enhancement when the data volume scales up. The proposed approach is verified with experiments using real-world data.

The Second International Workshop on Service and Cloud Based Data Integration (SCDI 2014) intends to bring researchers, practitioners and vendors together to discuss and share ideas and experiences. It fosters novel models, methodologies, and solution patterns that address the data integration issue and fit in the service and cloud based settings. This workshop will focus on the use of service and/or cloud based technologies to meet the new data integration challenges that are not well served by the current approaches.

With the growth and the diversification of web data, as an approach to realize data mashup service combination cannot satisfy the flexibility and simplicity of data integration. In this paper we proposes service hyperlink to save the relationship between two data services and utilize reusing the service hyperlinks to combine the data services flexibly and easily. First, we develop an algorithm to quickly and efficiently generate the service hyperlink. Then we publish the links and reuse them to make mashup more flexible and easy-to-use. We illustrate the service hyperlink by the scenarios of criminal investigation for police officers.