With the rapid development of service computing and software technologies, it is necessary to correctly and efficiently classify web services to promote their discovery and application. The existing... Show More

With the rapid development of service computing and software technologies, it is necessary to correctly and efficiently classify web services to promote their discovery and application. The existing service classification methods based on heterogeneous information networks (HIN) achieve better classification performance. However, such methods use negative sampling to randomly select nodes and do not learn the underlying distribution to obtain a robust representation of the nodes. This paper proposes a web services classification method based on HIN and generative adversarial networks (GAN) named SC-GAN. The authors first construct a HIN using the structural relationships between web services and their attribute information. After obtaining the feature embedding of the services based on meta-path random walks, a relationship-aware GAN model is input for adversarial training to obtain high-quality negative samples for optimizing the embedding. Experimental results on real datasets show that SC-GAN improves classification accuracy significantly over state-of-the-art methods.

The main function of recommendation systems is to help users select satisfactory services from many services. Existing recommendation systems usually need to conduct a questionnaire survey of the... Show More

The main function of recommendation systems is to help users select satisfactory services from many services. Existing recommendation systems usually need to conduct a questionnaire survey of the user or obtain the user's third-party information in the case of cold start users; this operation often infringes on the user's privacy. This article is aimed at providing accurate recommendations for cold start users without infringement on user privacy. Therefore, in response to this problem, this manuscript per the authors proposes a recommendation algorithm based on Hofstede's cultural dimensions theory. The algorithm uses Hofstede's cultural dimensions theory to establish a connection between two cold start users, thus ensuring the stability of QoS prediction accuracy. Then, the prediction results and the dynamic combination of the matrix factorization algorithm are used to obtain a more accurate prediction. The verification results on the real dataset WS-Dream show that the prediction algorithm proposed in this paper effectively alleviates the user cold start problem.

Clothes-changing person re-identification is a hot topic in the current academic circles. Most of the current methods assume that the clothes of a person will not change in a short period of time... Show More

Clothes-changing person re-identification is a hot topic in the current academic circles. Most of the current methods assume that the clothes of a person will not change in a short period of time, but they are not applicable when people change clothes. Based on this situation, this paper proposes a dual-branch network for clothes-changing person re-identification that integrates a two-level attention mechanism and captures and aggregates fine-grained person semantic information in channels and spaces through a two-level attention mechanism and suppresses the sensitivity of the network to clothing features by training the clothing classification branch. The method does not use auxiliary means such as human skeletons, and the complexity of the model is greatly reduced compared with most methods. This paper conducts experiments on the popular clothes-changing person re-identification dataset PRCC and a very large-scale cross-spatial-temporal dataset (LaST). The experimental results show that the method in this paper is more advanced than the existing methods.

In the steelmaking process, ensuring stable and reliable furnace plays a vital role for guaranteeing production quality of steel products. Traditional methods for detecting furnace anomalies in... Show More

In the steelmaking process, ensuring stable and reliable furnace plays a vital role for guaranteeing production quality of steel products. Traditional methods for detecting furnace anomalies in blast furnaces rely on operator judgment models built upon expert knowledge that can be limited by human experience. Moreover, data generated in blast furnace ironmaking process can be multidimensional, non-Gaussian distributed, and periodical, which can be easily affected by environmental and human factors and thus resulting in low accuracy of anomaly detection. Therefore, an online intelligent framework for detecting furnace anomalies is in high need. In this paper, the authors propose a novel anomaly detection method based on a furnace condition parameter-characterization model, a mining model of periodic patterns in the ironmaking process, and a multi-domain adaptive anomaly detection algorithm. They conduct extensive numerical analysis based on real-world production datasets as well to evaluate the effectiveness and accuracy of the method.

Dike health inspection is crucial in river channel regulating. The conventional manual collapse inspection is inefficient and costly so that the unmanned aerial vehicle (UAV)-based inspection has... Show More

Dike health inspection is crucial in river channel regulating. The conventional manual collapse inspection is inefficient and costly so that the unmanned aerial vehicle (UAV)-based inspection has been widely applied. However, the existing vision-based defect detection methods face challenges, such as lack of defect sample data and closed specified data sets. To address them, a defect detection method based on improved YOLOv5 recognition combined with image morphology processing is proposed for dike health inspection with zero defect samples. Specifically, the coordinate attention mechanism is introduced in YOLOv5 model to improve recognition capability for dikes. Also, a rotating bounding box target detection is designed for arbitrary orientation of dikes under UAV view, due to ineffective horizontal bounding box detection. Furthermore, for suspected defect locating efficiency promotion, the specific recognized area of the dike is isolated in the image morphology process. The results show that the proposed method outperforms the traditional Yolov5 algorithm on recall rate, F1, and mAP.

This study analyses the main techniques for constructing a three-dimensional (3D) digital core, and summarizes the advantages and disadvantages of each reconstruction technique. The direction for... Show More

This study analyses the main techniques for constructing a three-dimensional (3D) digital core, and summarizes the advantages and disadvantages of each reconstruction technique. The direction for the development of reconstruction technology for 3D digital cores is also proposed. The integration of multiscale, multicomponent, and nanoscale technology under different resolutions will play a significant role in future research efforts surrounding the physical properties of 3D digital cores. This study also introduces the main evaluation methods of digital cores, while analyzing and comparing the applicable scenarios of each method. The results of the pore statistics method, based on the seed filling algorithm, can be used as a new and effective evaluation method for the 3D reconstruction of digital cores.

In the digital era, enterprises have established online innovation communities to attract customers to participate. Presented in this study is user interactions within these communities using social... Show More

In the digital era, enterprises have established online innovation communities to attract customers to participate. Presented in this study is user interactions within these communities using social network analysis. By identifying distinct subgroups within the network and comparing the user interactions among these subgroups, this research aims to identify the group diversity of online interactions. The findings indicate that dialogists are more willing to interact and hold a favorable network position, followed by questioners, while answerers have the lowest level of interaction. User subgroups are identified using k-core analysis. The higher the value of the core k, the more interactions between users in the k-core subgroup and the better the network position. By combining both ego-centered and group dimensions of online interaction characteristics, this paper also outlines an investigation into an empirical study on the influence of user interactions on community recognition. The results confirm heterogeneous effects among different subgroups.

With the increasing awareness of personal health, personal health data management has become an important part of people's lives. Smart wearable devices (SWDs) collect people's personal health data... Show More

With the increasing awareness of personal health, personal health data management has become an important part of people's lives. Smart wearable devices (SWDs) collect people's personal health data, and then store the data on cloud. Authorized entities access the data to provide personalized health services. However, these personal health data contain a large amount of sensitive information, which may pose a significant threat to people's lives and property. To address this, this paper proposes a privacy-preserving solution. SWD data is encrypted, and secure indexes are created using Bloom filter and 0-1 encoding. Encrypted data and indexes are stored in a semi-trusted cloud. Only authorized entities can access the ciphertexts, ensuring secure personalized health management. Extensive experiments validate the scheme's efficiency in index construction, query token generation, and ciphertext search. Security analysis confirms no external entity, including the cloud, gains additional information during retrieval.

This article proposes a recommendation model based on self-attention mechanism and DeepFM service, the model is SelfA-DeepFM. The method firstly constructs the service network with DTc-LDA model to... Show More

This article proposes a recommendation model based on self-attention mechanism and DeepFM service, the model is SelfA-DeepFM. The method firstly constructs the service network with DTc-LDA model to mine the potential relationship between Mashup and API, which not only fully considers the text attributes but also combines the network structure information to effectively mitigate the sparsity of the service data. Secondly, service clustering to obtain numerical feature similarities. Finally, the self-attention mechanism is used to capture the different importance of feature interactions, and the DeepFM model is used to mine the complex interaction information between multidimensional features to predict and rank the quality score of API services to recommend suitable APIs. To verify the performance of the model, the authors use the real data crawled from the ProgrammableWeb platform to conduct multiple groups of experiments. The experimental results show that the model significantly improves the accuracy of service recommendation.