查看更多>>摘要：Utilization data is a kind of time-series data that consists of the proportion of the system's busy time in a fixed time interval. Utilization data can indicate the status of servers in a computer system, such as CPU utilization. Unfortunately, estimating model parameters from utilization data is challenging due to the inability to obtain the exact job arrival time and service time. Moreover, the maximum likelihood estimation(MLE) method tends to be sensitive, which may cause an overfitting problem. In this paper, a hierarchical Bayes (HB) based approach is proposed to estimate the parameters of queueing systems from utilization data. Specifically, a time non-homogeneous queueing system M_t/M/1/K whose job arrival follows a non-homogeneous Poisson process (NHPP) is supposed. Then, a series of homogeneous Poisson processes (HPP) is approximated to simplify the NHPP. Finally, the HB method is applied to estimate parameters of the M_t/M/1/K to address the sensitive issue of the MLE method. In numerical experiments, the effectiveness of the proposed HB-based approach with CPU utilization data is validated. In addition, the statistical properties of estimated parameters with MLE and HB are also studied in experiments.
查看更多>>摘要：Personal education information can be obtained online, which is convenient for inquiries, but at the same time brings the threat of privacy leakage. At present, everyone's education information is managed by a centralized organization. Centralized management has some security and privacy risks. Once the central management agency has a security problem, it may lead to the disclosure of everyone's personal information. Therefore, a privacy protection scheme for personal education information based on attribute-based encryption and homomorphic encryption on blockchain is proposed. Blockchain is a decentralized system and its anti-tampering features are very suitable for protecting data security. In addition, an attribute-based encryption method that performs fine-grained access control to the data content is proposed, which can decrypt the corresponding content according to the attributes held by the user. The use of homomorphic encryption can provide the required query results without knowing the plaintext. Access control, as well as key management, is implemented on Ethereum with smart contracts. Finally, security analysis and an example of privacy protection of personal education information are presented, and time performance is analyzed.
查看更多>>摘要：Sensitivity analysis plays a critical role in quantifying uncertainty in the design of computer systems. In particular, a variance-based global sensitivity analysis is often used to rank the importance of input factors based on their contribution to the variance of the output measure of interest. The variance-based sensitivity analysis is sampling-based and therefore usually applies simulation methods such as Monte Carlo simulation. That means the traditional methods for variance-based sensitivity analysis based on simulation do not need the analytic structure of the model to be analyzed. However, the simulation usually needs a huge number of realizations to obtain stable results, which incurs an undesired high computational cost. In this paper, we present an analytic approach to compute the variance-based sensitivity based on moment approximation. More specifically, we formulate the output measure of continuous-time Markov chains (CTMCs) and investigate the relationship between input parameters and output measure through variance-based sensitivity analysis. The numerical results showing the main effects of model parameters in both parallel and series system configurations indicate that a component's effect on the uncertainty in system reliability depends largely on the system structure.
查看更多>>摘要：The research of mobile health application interventions has attracted much interest from practitioners and scholars. There is a relative paucity of research investigating the factors influencing consumers to use a health chatbot for weight and health management. This study utilized the extended unified theory of acceptance and use of technology (UTAUT2) model as the theoretical basis and extended it with personal innovativeness and network externality to investigate the predominant factors affecting one's intention to use a health chatbot. Materials and Methods: An online survey was carried out for people aged≥20 years throughout Taiwan from 23 November to 30 December, 2019. Structural equation modeling (SEM) was used to test the hypotheses. Results: In total, 415 responses were analyzed. Our proposed model explained 87.1% of the variance in behavioral intention. Among eight factors, five of the factors were found to be the significant predictors of intention to use a health chatbot. Gender and experience were seen to exert a moderating effect on some of the relationships hypothesized in our research model, whereas education, chronic conditions, BMI. and age did not play a significant role. This study provides academics, health professionals, and practitioners with insights into the factors influencing the acceptance and use of a health chatbot. In the future, researchers could extend the model to investigate the effects of user intention on actual use behavior.
查看更多>>摘要：With the rapid development of the Internet of Things in recent years, there has been a dramatic increase in terminal units and new computationally and data-demanding applications. A terminal unit uploads data to the cloud server, which will be transmitted back to the terminal unit after certain operations. However, such a traditional cloud service is troubled by growing latency. Mobile edge computing emerges in such an environment. A short distance between the edge network and end-users mitigates this problem. However, the edge network has finite resources, making it impossible to deliver all service caching requests. To this end, a strategy is required to selectively cache services on the edge cloud. This study simulates the selection of edge services with a multi-armed bandit model and conducts a comparative study to analyze the impact that different algorithms have on performance.
查看更多>>摘要：Research on reliability of a power grid based on complex network theory has become a hot topic as the complexity of power grids increases with the continuous development of complex network theory. The research status of reliable operation of a power grid has been discussed by scholars from the perspectives of reliability, vulnerability, resilience and robustness. However, most studies are presented only from a single perspective. Thus, this paper comprehensively reviews power grid reliability research status from the perspective of key elements of identification and robustness analysis, which represent perturbation limit and tolerance to perturbation, respectively. Firstly, the basic theory and common evaluation indexes of complex network, as well as the influence of key elements and robustness on the reliable operation of power grid are introduced. Secondly, the current status of key element identification in power grids, the existing power grid elements importance ranking algorithms, its advantages and disadvantage, and evaluation criteria are discussed. Then, the present status of robustness analysis of power grids and the issues to be further explored are presented. Finally, the summary and outlook are given.
查看更多>>摘要：Recently, the internet has become the first destination as a recruitment market, which has increased the number of job offers and resumes online. Recommendation systems are proposed to help users filter this massive amount of information, selecting the best candidate or the relevant offer. Processing the content of the documents correctly not only can reduce the matching complexity but also improve the recommender performance. This paper presents a semantic-based information extraction process, which intelligently and automatically extracts domain entities. The extracted entities are inter-linked to build domain context utilizing domain ontology covering the most significant and common parts of job offers/resumes. Moreover, the extracted information is structured in RDF triples delivering a semantic and unified presentation of documents data. The used ontology is dynamically enriched with both domain instances and relations to keep up with the constant change of the relevant data. We evaluate our system using various experiments on data from real-world recruitment documents. Our test results show that our approach can achieve a precision value of more than 90% in extracting domain-specific information.
查看更多>>摘要：Pneumonia is a pandemic that needs to be diagnosed early to prevent miserable deaths. The diagnosis of the condition takes longer using traditional procedures. The diagnosis of a chest X-ray image has been improved since the development of medical imaging technologies. Different layer numbers are used to create advanced learning methods. In pediatric patients, X-ray pictures of the chest are carefully chosen. As part of the patient's overall health treatment, an X-ray image of the tissue is obtained. Convolution Neural Networks (CNN) also contains neurons with observable weights and biases. Each neuron takes a specific input, creates a dot product, and then moves out of line voluntarily. To acquire the biggest contour in the sliced images, contract limited adaptive histogram equalization is being utilized in this paper. Last but not least, images are included to CNN models. On both sides of the image, the median difference of the variable histogram analyzes on the chest X-ray image is taken, and then it is trimmed. By comparing the performance metrics for all in-depth learning methods, the proposal is based on an integrated method utilizing image processing and a separate VGG-16 and VGG-19 which are distributed along with the InceptionResNetV2 research model separately. Hence, three different Deep CNNs to automatically detect pneumonia in chest are used in this work.
NSTL