“…45 Therefore, some researchers have used the time series to predict the QoS values of the web services. [46][47][48] Moreover, some methods added the time-of-service invocation as a context to memory-based or model-based prediction methods to increase the accuracy of predictions and provide a context-aware method. For the first time, Zhang et al have used tensor factorization to provide a time-aware QoS prediction method.…”
The number of web services available on the internet has exploded, and as a result, the number of services with the same functionality has exploded as well. Therefore, selecting the best web service from functionally similar services is a critical task in the web service domain. The Quality of Service (QoS) is one of the most common criteria used to select the best web service. Collaborative filtering (CF) has been utilized in several studies to predict the values of QoS attributes of web services for each user in a personalized way. The QoS histories of other users are employed in these methods to predict the QoS values of the active user. Although these methods function well and produce acceptable prediction results, the accuracy of their predictions can be harmed by incorrect data provided by untrustworthy users. In this study, we propose a new model that reduces the impact of unreliable user data, resulting in a trustworthy prediction. This model can be applied to any existing prediction method. In experiments, the proposed model was applied to seven known prediction methods. The results indicate that this model is able to eliminate the impact of unreliable users.
“…45 Therefore, some researchers have used the time series to predict the QoS values of the web services. [46][47][48] Moreover, some methods added the time-of-service invocation as a context to memory-based or model-based prediction methods to increase the accuracy of predictions and provide a context-aware method. For the first time, Zhang et al have used tensor factorization to provide a time-aware QoS prediction method.…”
The number of web services available on the internet has exploded, and as a result, the number of services with the same functionality has exploded as well. Therefore, selecting the best web service from functionally similar services is a critical task in the web service domain. The Quality of Service (QoS) is one of the most common criteria used to select the best web service. Collaborative filtering (CF) has been utilized in several studies to predict the values of QoS attributes of web services for each user in a personalized way. The QoS histories of other users are employed in these methods to predict the QoS values of the active user. Although these methods function well and produce acceptable prediction results, the accuracy of their predictions can be harmed by incorrect data provided by untrustworthy users. In this study, we propose a new model that reduces the impact of unreliable user data, resulting in a trustworthy prediction. This model can be applied to any existing prediction method. In experiments, the proposed model was applied to seven known prediction methods. The results indicate that this model is able to eliminate the impact of unreliable users.
“…This architecture design, therefore, allows us to customize every part of the system and keep them uncoupled, facilitating the system's maintenance and evolution. In addition, we use the Autoregressive Integrated Moving Average (ARIMA) prediction model [45], which is not used in any of the mentioned approaches. ARIMA works adequately with time series, making real-time predictions reliable.…”
The impressive evolution of the Internet of Things and the great amount of data flowing through the systems provide us with an inspiring scenario for Big Data analytics and advantageous real-time context-aware predictions and smart decision-making. However, this requires a scalable system for constant streaming processing, also provided with the ability of decision-making and action taking based on the performed predictions. This paper aims at proposing a scalable architecture to provide real-time context-aware actions based on predictive streaming processing of data as an evolution of a previously provided event-driven service-oriented architecture which already permitted the context-aware detection and notification of relevant data. For this purpose, we have defined and implemented a microservice-based architecture which provides real-time context-aware actions based on predictive streaming processing of data. As a result, our architecture has been enhanced twofold: on the one hand, the architecture has been supplied with reliable predictions through the use of predictive analytics and complex event processing techniques, which permit the notification of relevant context-aware information ahead of time. On the other, it has been refactored towards a microservice architecture pattern, highly improving its maintenance and evolution. The architecture performance has been evaluated with an air quality case study.
“…Most of the QoS features are dynamic. Dynamic features are attributes that do not have fixed values and their values depend on various factors such as network infrastructure, time of invocation, and users' location 8 . Response time, throughput, and availability are dynamic features.…”
Section: Introductionmentioning
confidence: 99%
“…In model‐based methods, using a user‐service invocation matrix, a pattern is designed, and then by learning from these training data, the model can predict unknown QoS values. Some papers have used clustering, matrix factorization (MF), time series, and machine learning techniques for model‐based prediction of QoS 8,17,24‐26 …”
Quality of Service (QoS) of Web services plays an essential role in selecting Web services by consumers. The dynamic QoS attributes of Web services have different values for different users. Therefore, the value of many Web services' QoS features for many users are undetermined, and these values should be predicted. The collaborative filtering (CF) method is one of the most successful approaches to predict these values. CF‐based methods use the QoS values contributed by the other users for prediction and, consequently, the values contributed by unreliable users can decrease the accuracy of prediction. To utilize the reputation of users can be regarded as one of the conventional approaches to overcome this problem. In this paper, we have defined a concept called regional reputation that represents the reputation of a user for users in each geographical region. Regional reputation has been achieved with the combination of the location information of the users and their reputation. Subsequently, by combining this concept with the matrix factorization, we have proposed a prediction method called regional reputation‐based matrix factorization. This approach has been able to improve the accuracy of prediction and be more persistent to the data contributed by unreliable users.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.