Application of Bayesian Networks for Autonomic Network Management
Application of Bayesian Networks for Autonomic Network Management
复制标题
DOI：
10.1007/s10922-013-9289-x
复制
发表时间：
2014-04
影响因子：
3.6
通讯作者：
A. Bashar;G. Parr;S. McClean;B. Scotney;D. Nauck
中科院分区：
文献类型：
--
作者：
A. Bashar;G. Parr;S. McClean;B. Scotney;D. Nauck
电信网络在技术、基础设施和支持的业务方面不断发展，为网络管理人员提供高效的网络管理系统（NMS）提出了新的挑战。对当前网络，更具体地说是下一代网络（NGN）的自动化和高效管理的需求是本研究的主题。本文详细描述了当前网络环境下的管理挑战，然后列出了高效NMS所需的特征和特征。然后提出有必要应用人工智能（AI）和机器学习（ML）方法来增强和自动化NMS的功能。这项工作的第一个贡献是对应用于NM领域的AI和ML方法的全面调查。这项工作的第二个贡献是，它提供了推理和证据来支持选择贝叶斯网络（BN）作为基于ml的NMS的可行解决方案。这项工作的最后贡献是，它提出并实现了基于BN方法的三种新颖的NM解决方案，即基于BN的准入控制（BNAC），基于BN的分布式准入控制（BNDAC）和基于BN的智能交通工程（BNITE），以及支持所提议框架的算法描述。
The ever evolving telecommunication networks in terms of their technology, infrastructure, and supported services have always posed challenges to the network managers to come up with an efficient Network Management System (NMS) for effective network management. The need for automated and efficient management of the current networks, more specifically the Next Generation Network (NGN), is the subject addressed in this research. A detailed description of the management challenges in the context of current networks is presented and then this work enlists the desired features and characteristics of an efficient NMS. It then proposes that there is a need to apply Artificial Intelligence (AI) and Machine Learning (ML) approaches for enhancing and automating the functions of NMS. The first contribution of this work is a comprehensive survey of the AI and ML approaches applied to the domain of NM. The second contribution of this work is that it presents the reasoning and evidence to support the choice of Bayesian Networks (BN) as a viable solution for ML-based NMS. The final contribution of this work is that it proposes and implements three novel NM solutions based on the BN approach, namely BN-based Admission Control (BNAC), BN-based Distributed Admission Control (BNDAC) and BN-based Intelligent Traffic Engineering (BNITE), along with the description of algorithms underpinning the proposed framework.