Special issue on big data intelligence in communication systems

With the development of communication systems,e.g., mobile, Internet and telecommunication systems, a large volume of data in various formats is collected much easier than before. As thus, it is more common than ever to deal with large datasets to realize the intelligence of these public communication systems. However, with the advent of wireless sensor networks, social networks, smart tools, and the Internet, data with heterogeneity, high-volume and low-quality is generated rapidly, which makes it harder to gather, transfer, store, and fuse the data as well as analyze the behaviors of the systems,e.g., security anomalies detection or future demands prediction. Fortunately, the advances in high-performance computing power and big data analysis along with intelligent techniques, called Big Data Intelligence, can handle the huge stream of operational data from these systems and can perform analytical processing to promote its usability, effectiveness, and efficiency, which offers us enormous opportunities and transformative potential for predictive services and intelligent decisions. Extensive attention has been captured recently to improve operations and managements for communication systems by employing the intelligence of big data. It not only provides a comprehensive understanding and a promising decision-making framework based on massive data but also gives opportunities to design novel algorithms and platforms for its management and analytics.

Therefore, it is high time to call for contributions to further stimulate the continuing efforts to enable such an integrated framework, some fundamental issues are required to be addressed to provide the necessary bridge between Big Data Intelligence and Communication System. This special issue attracted numerous submissions, out of which eight papers have been accepted to be published. These accepted papers are expected to highlight some of the important aspects outlined above. It is the pleasure for the guest editorial team to introduce these papers as follows.

The paper entitled “Design and modeling of survivable network planning for software-defined data center networks in smart city” by Y. Peng et al proposes a new software-defined networking based survivable network planning model. In this model, the traditional elastic network planning is defined as a mixed-integer optimization problem, which aims to seek the minimum number of unprotected nodes. Correspondingly, three novel methods, namely, the simulated annealing based K-means clustering, Lagrangian relaxation, and greedy routing, are designed to obtain the feasible solutions.

The paper entitled “Analysis of instantaneous availability of communication system based on the influence of support equipment” by Y. Yang et al designs a novel instantaneous availability (IA) model for communication systems. Inspired by the queuing theory and Markov process, it mainly focuses on studying two typical fault modes in communication systems and gives the corresponding solutions. On one hand, an IA model that considers the repair equipment failure in the maintenance of electronic units is proposed. On the other hand, the enhanced IA model that explores the queuing problem of multiple electronic units is further developed. Moreover, the M/G/1 queuing system is employed to identify the distribution functions for the waiting and service time.

In the paper entitled “An intelligent efficient scheduling algorithm for big data in communication systems” by F. Bu, proposes a reinforcement learning-based scheduling algorithm to increase the utilization and reduce the energy consumption of the processors for large-scale data analysis. The main contributions are as follows. First, a reinforcement learning model is developed to identify a suitable dynamic voltage. After that, a frequency scaling technique which can configure the voltage and frequency is proposed based on the state of current system. Thus, the utilization can be improved and the energy consumption can be optimized effectively. In addition, the parameters of the reinforcement learning model are trained by a newly designed learning algorithm.

The paper entitled “Knowledge fusion based on the group argumentation theory in Web 2.0 environment” by X. Chen et al aims to formalize the structured knowledge representation in Web 2.0 environment. Thus, in this paper, a knowledge element model is proposed, where the cognition viewpoints of group members at something are abstracted into different knowledge element versions. Regarding the attributes that are divergent among all the versions of knowledge element, the corresponding group argumentation tasks are developed. Thus, when employing this new proposed model, all attributes belonging to the knowledge element can be determined promisingly. As a result, the attribute set for the knowledge element is formulated, and the knowledge fusion is obtained.

The paper entitled “Research on the dynamic effect of the intelligent urban experience to the tourists' two-way internet word-of-mouth” by T. Ma et al studies the relations among tourism enterprises from Gu networking, and the relations among users with negative word-of-mouth. Specially, quantitative analysis on the evolution model in tourism enterprises when there are external competitions investigate the negative reputation of users. Also, the users with positive and negative word-of-mouth can be identified. Thus, the corresponding countermeasures that the tourism enterprises should take can be given. Moreover, the essence of tourism enterprises and the relations among negative word-of-mouth competitor are defined by analyzing the evolution of phase diagram.

The paper entitled “An effective copyright-protected content delivery scheme for P2P file sharing networks” by M. Su et al presents a novel copyright-protected content delivery scheme (CPP) in spired by the superiority of piece-level stochastic encryption. In the proposed CPP, when accessing to the plaintext content, peers usually have to decode the decryption key sequences. Since different CPP has different sequence of the decryption keys, the unauthorized peers have no access to the contents even though corresponding colluders share the keys with them. Usually, the hash modification is able to make the peers to check each piece's correction after it is downloaded, thus it can prevent the propagation of poisoned pieces. Moreover, during the download process, neither massive state maintenance nor frequent user authentication is introduced. In addition, peers purchase after downloading, which can protect their benefits even if the system is quitted before paying.

The paper entitled “Automatic fast double KNN classification algorithm based on ACC and hierarchical clustering for big data” by H. Li et al designs a new KNN classification algorithm for large-scale data. In the proposed method, the cluster centers are determined automatically for training the KNN classification. Specially, clustering methods are employed to divide the data samples into several parts. In the testing process, the clusters that have the nearest distances to the testing samples are selected as the new training samples. After that, each new sample is further handled by the hierarchical clustering. In this way, the computational consumption and time complexity can be reduced significantly.

In the paper entitled “Modified pyramid dual tree direction filter-based image denoising via curvature scale and nonlocal mean multigrade remnant filter” by L. Teng et al, a modified pyramid dual tree direction filter for image denoising is proposed. First, the curvature scale model, which employs robust Bayes least square to calculate pyramid dual tree direction filter coefficients, is applied for noised images. After that, the inverse transformation of the filter is employed to reconstruct the initial denoised images. Finally, the nonlocal mean multigrade remnant filter is used to filter the initial denoised images and thus the processed images are achieved.