Deviation-based neighborhood model for context-aware QoS prediction of cloud and IoT services
Introduction
With the rapid development and widespread deployment of cloud computing and Internet of Things (IoT) technologies, more and more homogeneous services emerge on the Internet [1]. In particular, the IoT will supply and enable us to access all of real-world facilities, resources and information through the Internet [2]. If all of these resources are provided to users in a service-oriented manner, we will confront the challenges of the explosion of services. Given a set of functionality-similar services, how to select a service which best matches a user’s needs has become an important practical issue [3]. In the traditional approaches, client-based service evaluation was often used to provide users with real-time measurement of the quality of services [4], [5], [6]. With distinguishable quality values of candidate services, users can now take a decision on choosing appropriate services [7]. However, this treatment is not realistic in large-scale service systems both for service providers or users. On the one hand, service providers need to deploy a large number of distributed software sensors to monitor the quality information of services, and thus lead to a lot of economic costs. On the other hand, it will take much time for users to experience the unique differences of QoS for every candidate services. Consequently, how to obtain personalized QoS of cloud/IoT services and assist users selecting appropriate services remains as a challenging issue [8], [9].
Recently, researchers pursuit solution for this problem by drawing lessons from the recommender systems [10], [11], [12]. Assume there exists a centralized repository which can collect the QoS data of users in the use of cloud and IoT services. Based on the principle that similar users (or services) tend to observe similar quality scores on the same service (or user), collaborative QoS forecasting models are built to predict the unknown quality values given active users and services. This treatment exploits crowd intelligence to aid the QoS assessment and avoid intuitive data measurement, thereby save time and economic costs for both service providers and users. Consequently, collaborative QoS prediction has become an increasingly active and highly problem-rich research area, and many works have been carried out to address this topic based on either neighborhood-based collaborative filtering (CF) [10] or matrix-factorization [11], [12].
With respect to the collaborative QoS prediction, the commonly used methods are neighborhood-based collaborative filtering (CF) [13] and matrix-factorization [14]. The advantages of neighborhood-based CF are simplicity, justifiability and efficiency [15]. However, these models are not justified by a formal model [16]. Moreover, heterogeneous similarity metrics and sparsity-sensitive problem make these models not robust and scalable enough. In contrast, matrix factorization approaches comprise an alternative approach to CF with the more holistic goal to uncover latent features from user–service usage data [16] thus can alleviate the problem of sensitivity to sparse data. Since matrix-factorization can be presented as a formal optimization problem and solved by machine learning methods, it provides attractive accuracy and scalability for QoS prediction. However, matrix factorization is always uncertain, where the learned latent factors are unexplainable, thus resulting in difficulty as explain the predicting results for users. It is very important that recommender systems provide explanations for their recommendations so that users can consider and trust them [17]. By the same token, users certainly hope to be able to get believable explanations for the QoS predictions provided by a cloud service recommendation system.
Consequently, a spontaneous concern may be questioned whether it can develop more accurate neighborhood-based models which overcome existing difficulties, and achieve more accurate prediction than matrix factorization. It would be a better solution for the task of QoS prediction. Inspired by this, this paper proposes learning neighborhood-based models for personalized QoS prediction of cloud/IoT services. We define formal deviation-based neighborhood models for QoS prediction and give an algorithmic framework which permits an efficient global optimization scheme for learning the models’ parameters. The proposed models obtain baseline estimates for QoS prediction using deviations of users and services, further exploit neighborhood components and contextual information to smooth and enhance the prediction accuracy. Experimental results show that our model can overcome existing difficulties and then perform superior to the-state-of-art prediction methods.
The remainder of the paper is organized as follows. Section 2 reviews some existing works that are most relevant to ours. Section 3 gives details of proposed deviation-based neighborhood prediction models. The formal foundation of the models is presented. Components for deviation-based baseline estimation and the component of neighborhood tier is given. Meanwhile, a unified learning algorithm is presented to optimize the parameters of the models. We evaluate the effectiveness of proposed methods via a set of experiments on real QoS data in Section 4 and make conclusions in Section 5.
Section snippets
Related works
Collaborative QoS prediction mainly borrows ideas from the researches of recommender systems [8], [10], [11], [12], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], where various collaborative filtering models are exploited to prediction unknown QoS values based on the QoS usage data in service-oriented systems. In this section, we shortly introduce some existing works that are most relevant to ours.
Deviation-based neighborhood model for QoS prediction
To cope with existing drawbacks of CF-based prediction methods, we suggest using machine learning techniques to build neighborhood model for QoS prediction. New models allow an efficient global optimization scheme and exploit different baseline estimate components to improve prediction accuracy. To distinguish users from services, we take different indexing letters: for users , , and for services , . The notation indicates a known quality-score observed by user on service and the
Datasets
To evaluate the QoS prediction performance, we use WS-DREAM dataset,1 a large-scale dataset collected and maintained by Zheng et al. [8], [39]. The dataset consists of a total of 1,974,675 real-world web service invocation results are collected from 339 users on 5, 825 real-world web services via PlanetLab platform,2 which is a global research network that supports the development of new network services. Users are linked to
Conclusion and future works
Guided by the principles embodied in collaborative filtering and machine learning, we propose deviation-based neighborhood models for making collaborative QoS-prediction of cloud/IoT services. The framework offers an efficient global optimization scheme then enables robust and accurate prediction results. Also, location information is utilized to strengthen the predicting ability of the proposed approaches. In addition, it preserves the explainability for the QoS-prediction tasks which would be
Acknowledgments
This work is supported by the Special Funds for Middle-aged and Young Core Instructor Training Program of Yunnan University, the Applied Basic Research Project of Yunnan Province (2013FB009, 2014FA023), China Postdoctoral Science Foundation funded project (2016M592721), Program for Excellent Young Talents of Yunnan University (No. XT412003), and the National Natural Science Foundation of China (61472345, 61402398, 61272158, 61562090). The authors are grateful to reviewers for their constructive
Hao Wu received the bachelor’s degree in computer science from Zhengzhou University, in 2001, master and Ph.D. degrees in computer science from Huazhong University of Science and Technology in 2004 and 2007, respectively.
Now, he is an associate professor at School of Information Science and Engineering, Yunnan University, China. He has published more than fifty papers in peer-reviewed international journals and conferences. His research interests include service computing, information filtering
References (43)
- et al.
Personalized QoS prediction of cloud services via learning neighborhood-based model
- et al.
Learning from explanations in recommender systems
Inform. Sci.
(2014) - et al.
Efficient web service QoS prediction using local neighborhood matrix factorization
Eng. Appl. Artif. Intell.
(2015) - et al.
Colbar: A collaborative location-based regularization framework for qos prediction
Inform. Sci.
(2014) - et al.
Another look at measures of forecast accuracy
Int. J. Forecast.
(2006) - et al.
A view of cloud computing
Commun. ACM
(2010) - et al.
Processing distributed Internet of things data in clouds
IEEE Cloud Comput.
(2015) - et al.
Cloudgenius: A hybrid decision support method for automating the migration of web application clusters to public clouds
IEEE Trans. Comput.
(2015) - G. Wu, J. Wei, X. Qiao, L. Li, A Bayesian network based QoS assessment model for web services, in: 2007 IEEE...
- et al.
On testing and evaluating service-oriented software
IEEE Comput.
(2008)
Workload prediction using ARIMA model and its impact on cloud applications QoS
IEEE Trans. Cloud Comput.
An infrastructure service recommendation system for cloud applications with real-time QoS requirement constraints
IEEE Syst. J.
Collaborative web service QoS prediction via neighborhood integrated matrix factorization
IEEE Trans. Serv. Comput.
Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions
IEEE Trans. Knowl. Data Eng.
Matrix factorization techniques for recommender systems
IEEE Comput.
A comprehensive survey of neighborhood-based recommendation methods
Factor in the neighbors: Scalable and accurate collaborative filtering
ACM Trans. Knowl. Discov. Data
QoS-aware web service recommendation by collaborative filtering
IEEE Trans. Serv. Comput.
Cited by (30)
QoS prediction for web services in cloud environments based on swarm intelligence search
2023, Knowledge-Based SystemsCitation Excerpt :To improve the PCC, Su et al. [15] considered data credibility problem, and identified the trustworthy similar users by combining the user reputation and the similarity between each other. In Ref. [8], the deviation-based neighborhood model was proposed to give a baseline estimate for QoS prediction by considering the QoS deviations of services and users, and then neighborhood-based CF was used to contribute the fine-grained adjustment of the predictions. To further increase the prediction accuracy, not only QoS records, but much supplementary information is also employed to conduct neighbor selecting from multiple perspectives.
Modeling feature interactions for context-aware QoS prediction of IoT services
2022, Future Generation Computer SystemsCitation Excerpt :Then, we report and discuss the experimental results. Currently, as there is no public dataset for IoT QoS prediction, almost all research literature on IoT QoS prediction published in reputable journals and conference (such as TSC [18,38,39], FGCS [17], IoT-J [19] and Percom [24]) use the dataset released by Zheng et al. [40] to perform experiment analysis, which consists of a matrix of the response time and throughput of 339 users for 5,825 web services. Therefore, we follow these work to measure the performance of the proposed method with this dataset, which can be freely downloaded from the website.1
Machine learning techniques in emerging cloud computing integrated paradigms: A survey and taxonomy
2022, Journal of Network and Computer ApplicationsCitation Excerpt :To predict the missing QoS values, the kernel machine learning algorithm and Kernel Least Mean Square (KLMS) algorithm are used to investigate the hidden connections between all known QoS data and the relevant QoS data with the best similarity. Wu et al. (2017) proposed context-aware QoS prediction using crowd intelligence. This utilized the Gradient Descent (GD) algorithm in a deviation-based neighborhood model.
A two-stage location-sensitive and user preference-aware recommendation system[Formula presented]
2022, Expert Systems with ApplicationsA survey on machine learning in Internet of Things: Algorithms, strategies, and applications
2020, Internet of Things (Netherlands)Optimal Fitness Aware Cloud Service Composition using an Adaptive Genotypes Evolution based Genetic Algorithm
2019, Future Generation Computer SystemsCitation Excerpt :Faheem et al. [33] proposed a novel dynamic clustering based energy efficient and quality-of-service (QoS)-aware routing algorithm, inspired by the real behavior of the bird mating optimization, to balance the data traffic and energy consumption load evenly among clusters in the smart grid. Wu et al. [34] developed a novel deviation-based neighborhood models for QoS prediction by taking advantages of crowd intelligence to obtain personalized quality of cloud/IoT services and assist users selecting appropriate services. Tomovic et al. [35] presented a software defined networking based architecture for virtualization of IoT networks, allowing virtualization of IoT resources in order to provide to users the intended service at the requested level of quality.
Hao Wu received the bachelor’s degree in computer science from Zhengzhou University, in 2001, master and Ph.D. degrees in computer science from Huazhong University of Science and Technology in 2004 and 2007, respectively.
Now, he is an associate professor at School of Information Science and Engineering, Yunnan University, China. He has published more than fifty papers in peer-reviewed international journals and conferences. His research interests include service computing, information filtering and recommender systems.
Kun Yue received the B.Sc., M.Sc., and Ph.D. degrees in computer science from Yun nan University (Kunming, China), Fudan University (Shanghai, China) and Yunnan University (Kunming, China) in 2001, 2004 and 2009, respectively. He is currently a professor at Yunnan University, Kunming, China. His current research interests mainly include massive data analysis and uncertainty in artificial intelligence.
Ching-Hsien (Robert) Hsu is a professor in department of computer science and information engineering at Chung Hua University, Taiwan; and distinguished chair professor in school of computer and communication engineering at Tianjin University of Technology, China. His research includes high performance computing, cloud computing, parallel and distributed systems, big data analytics, ubiquitous/pervasive computing and intelligence. He has published 200 papers in refereed journals, conference proceedings and book chapters in these areas. Dr. Hsu is the editor-in-chief of international journal of Grid and High Performance Computing, and international journal of Big Data Intelligence; and serving as editorial board for a number of prestigious journals, including IEEE Transactions on Service Computing, IEEE Transactions on Cloud Computing, etc. He has been acting as an author/co-author or an editor/co-editor of 10 books from Springer, IGI Global, World Scientific and McGraw–Hill. He has also edited a number of special issues at top journals, such as IEEE Transactions on Cloud Computing, IEEE Transactions on Services Computing, IEEE System Journal, Future Generation Computer Systems, Journal of Supercomputing, etc. Prof. Hsu was awarded eight times distinguished award for excellence in research and annual outstanding research award through 2005 to 2015 from Chung Hua University. He has been serving as executive committee of Taiwan Association of Cloud Computing (TACC) from 2008 to 2012; executive committee of the IEEE Technical Committee of Scalable Computing (2008–2012); IEEE Cloud Computing (2012–present); Dr. Hsu is an IEEE senior member;
Yiji Zhao is currently a Ph.D. candidate in School of Computer and Information Technology, Beijing Jiaotong University, Beijing, China. He former received the M.E. degree in computer science from School of Information Science and Engineering, Yunnan University. His research interests include complex network analysis, social network mining and recommendation algorithms.
Binbin Zhang was born in 1982. She received the Ph.D. degree in computer science from Peking University in 2011. She is currently a lecturer at Yunnan University. Her main research interests include virtualization and cloud computing.
Guoying Zhang is currently a postgraduate in School of Information Science and Engineering, Yunnan University. His current research interests include recommender systems, service computing and data mining.