改进遗传-狼群对节点序寻优的贝叶斯网络结构算法

贝叶斯网络是数据挖掘领域的一种重要方法。针对贝叶斯网络结构学习算法寻优效率低和易陷入局部最优的问题，提出一种基于改进的混合遗传-狼群对节点序寻优的贝叶斯网络结构学习算法。该算法首先利用深度优先搜索对最大支撑树的节点进行拓扑排序；然后利用动态变异及最优交叉算子构建适用于节点序寻优的改进捕食行为，引入动态参数因子来增强算法局部寻优能力；最后与K2算法结合得到最优的贝叶斯网络结构。用3种不同大小的标准网络数据集中进行实验，结果表明，该算法收敛到较优值，寻优效率高于其它同类优化算法。     

基于改进遗传算法的DBN结构自适应学习算法

针对动态贝叶斯网络(dynamic bayesian network,DBN)是NP困难问题,提出基于改进遗传算法的DBN结构自适应学习算法.该算法计算最大互信息和时序互信息完成DBN结构搜索空间的初始化.在此基础上设计改进遗传算法,引入评分标准差构建交叉概率和变异概率的自适应调节函数,以降低结构学习过程陷入局部最优解...     

基于贝叶斯网络的组合服务信任度评估方法

对组合服务信任度评估问题进行了研究。基于信任的不确定性和模糊性,给出了信任度的离散化表示方法;借助于贝叶斯网络的强大的推理能力,给出了组合服务的贝叶斯网络模型以及组合服务模型到贝叶斯网络的转换规则,利用贝叶斯网络的正向推理机制设计了基于贝叶斯网络模型的组合服务信任度评估方法。通过一个组合服务的买例,说明了组合服务的贝叶斯网络模型及其基于贝叶斯网络的组合服务信任度评估方法的使用方法。该评估方法基于信任的模糊性和不确定性,更符合信任的本质并易于实现。     

基于互信息的贝叶斯-案例检索特征选择模型

案例检索过程中案例集存在特征冗余,使得检索效率和准确度较低。传统的基于贝叶斯网络案例检索特征选择模型(BN-CBR)对先验知识利用效率不高,且不能有效选择消除冗余性的特征子集。构建基于互信息的贝叶斯-案例检索特征选择模型(MI-BNCBR),采用特征冗余度和互信息计算案例特征的综合权重,改善BN-CBR模型对先验知识利用效率不高的问题,其采用互信息方法可消除案例集中的冗余特征并得到最优特征子集,采用基于远端最近距离计算的K-D树方法进一步改善基于互信息的贝叶斯-案例检索的效率,并利用医学基准数据进行实验,结果表明所引入的方法有效地提高了案例检索的准确度和检索效率。     

基于代数决策图的贝叶斯网络参数简化技术

贝叶斯网络是一种进行不确定性知识表达和推理的有效工具,推理算法是贝叶斯网络研究的主要内容之一.目前,贝叶斯网络推理算法采用条件概率表(CPT)来存储贝叶斯网络中各节点的条件概率分布(CPD).CPT中的概率参数随父节点数目的增加呈指数增长,使得网络中概率参数急剧增加,降低了网络推理效率.为提高网络推理效率,本文提出采用代数逻辑图(ADD)取代CPT存储网络中各节点CPD的方法.结合有序二分决策图理论,分析并验证了ADD通过捕捉贝叶斯网络中父子节点之间的环境独立性来减少网络中的概率参数的原理,进而推导出了CPT到等价ADD转化的算法.最后,通过实例验证了ADD存储方式的有效性.结果表明,对于具有环境独立特性的贝叶斯网络,相对于CPT的存储方式,等价ADD存储方式可有效减少网络中的概率参数,为贝叶斯网络推理效率的提高提供一种有效手段.     

洗衣机交互信息定性定量表达与研究

目的基于用户特征研究洗衣机交互信息定性定量表达形式,为洗衣机交互信息表达提供指导。方法通过市场调研收集洗衣机交互信息,根据词性和Croft关联标记模式进行归纳总结,结合定性定量表达定义探讨洗衣机交互信息定性定量表达,并引入用户特征研究洗衣机交互信息中的程度、进度和度量类表达,探索用户特征对洗衣机交互信息定性定量表达的影响。结论获得了洗衣机交互信息定性定量表达的普遍规律及用户特征与洗衣机交互信息表达的关系,并通过实际项目进行了运用。     

基于局部图互信息最大化的异构图神经网络方法

针对现有的基于互信息最大化的异构图神经网络(HGNN)方法因图读出操作的单射限制、粗粒度的特征保留而无法适用于现实网络的问题,提出一种基于局部图互信息最大化的、无监督的异构图神经网络方法。该方法使用元路径对异构图中涉及到的语义关系进行建模,并利用图卷积模块和语义级别的注意力机制来捕获单个节点的局部表征。该方法通过最大化单个节点与局部子图间的互信息,有效地学习高阶节点表征。实验结果表明,该方法相比基于全局图互信息的方法,可以将数据集DBLP/IMDB上的节点分类任务的微值F1(micro-F1)提高大约3%/9%,同时将DBLP/IMDB上的节点聚类任务的调整兰德系数(ARI)提高约23%/46%。     

机器学习分布式网络传输异常数据智能检测方法

对于链路状态数据库的网络传输异常数据检测存在检测数据不完整、较为敏感、检测效率差的问题,提出基于机器学习的分布式网络传输异常数据智能检测方法,通过K最近邻分簇算法对分布式网络节点实施分簇,利用贝叶斯分类算法检测簇头是否出现异常;确定异常簇后,选取小波阈值降噪方法对异常簇内数据进行降噪处理,在此基础上,采用遗传算法检测降...     

基于Mamdani模糊推理的无线传感器网络可信簇头选举算法

为了解决无线传感器网络中恶意节点成为簇头而引起的层次路由安全问题,提出了一种基于Mamdani模糊推理的可信簇头选举算法(TCEM)。TCEM算法根据节点的行为表现,采用贝叶斯原理评估节点信任值,在此基础上,结合密集度及向心度,使用Mamdani模糊推理方法计算节点优越度,选择最优者作为簇头,从而实现簇头的可信选举。仿真实验结果表明,该算法能有效阻止恶意节点成为簇头,且在簇头合理分布、网络整体能效等方面均有良好表现。     

基于链路动态特征的自组网路由选择方法

针对在动态的移动ad hoc网络环境中失效路由的频繁重建和维护操作造成大量控制负载开销以及能量消耗的问题,提出了一种基于无线逻辑链路信息熵度量的路由稳定性计算方法.此算法通过网络运行过程中无线链路状态变化所表现出的不确定性特征来刻画网络节点的稳定程度,并在路由建立过程中选择相对稳定的无线移动节点作为数据传输的参与节点....     

基于种群生态学的竞争项目组合风险决策模型

在回顾了项目组合管理研究之后,将种群生态学理论引入项目组合风险管理.讨论了项目组合的已有研究,重新给出了项目组合的定义,并将其分为三种类型,揭示了项目组合与生物种群间的相似属性.构建了竞争型项目组合风险管理(CPPRM)决策模型,讨论了n=2时模型的解,分析了其稳定性条件及风险应对策略,并以n=3时的讨论论证了CPPRM决策模型的实际可操作性.     

A hybrid risks-informed approach for the selection of supplier portfolio

In conventional supplier selection approaches, cost consideration is usually emphasised and it renders a vulnerable supply chain with various risks. This article aims to develop a quantitative approach for modelling both supply chain operational risks and disruption risks to support decision-making with regard to order allocation and risk mitigation. We introduce two types of risk evaluation models: value-at-risk (VaR) and conditional value-at-risk (CVaR). Specifically, VaR is used to measure operational risks caused by improper selection and operations of a supplier portfolio to the stochastic demand, which may frequently occur but result in relatively small losses to supply chains; CVaR is used to evaluate disruption risks that are less frequent and tend to cause significant damage. After incorporating risk factors into a probability-based multi-criteria optimisation model, different methods and parameters are compared and tested to determine the factors that may influence the supplier selection process. Computational examples by simulation are presented to illustrate the approach and how decision-makers make trade-offs between costs and hybrid risks.     

A Bayesian network approach for fixture fault diagnosis in launch of the assembly process

Verification and correction of faults related to tooling design and tooling installation are important in the auto body assembly process launch. This paper introduces a Bayesian network (BN) approach for quick detection and localisation of assembly fixture faults based on the complete measurement data set. Optimal sensor placement for effective diagnosis of multiple faults, structure learning of the Bayesian network and the diagnostic procedure are incorporated in the proposed approach. The effective independence sensor placement method is used to reach the desired number of optimal sensor locations, which provide the concise and effective sensor nodes to build the diagnostic Bayesian network. A new algorithm based on conditional mutual information tests is put forward to learn the Bayesian network structure. The body side assembly case was used to illustrate the suggested method and the simulation analysis was performed to evaluate the effectiveness of the diagnostic network. The work demonstrated that the proposed methodology composes a feasible and powerful tool for fixture fault diagnosis in launch of the assembly process.     

Project Portfolio Ranking based on the Strategic Contribution Efficiency

Project portfolio management is proved to be remarkable for an enterprise to attain competences. When selecting project portfolios,enterprises are willing to know which one has the largest return on investment rate when realizing enterprise strategy. A methodology for project portfolio ranking by adopting the concept of Strategic Contribution Efficiency（ SCE） is proposed. SCE acts as the measure that demonstrates what degree a project portfolio would contribute to enterprise strategy at specific cost. Evaluation criteria are established according to the practical requirements of enterprises. Data Envelope Analysis（ DEA） is combined with fuzzy set to calculate SCE and the ranking,which takes project portfolio as a whole rather than evaluates individual projects separately and considers information imprecision at the same time. At last,a numerical example is illustrated the specific process of project portfolio ranking based on SCE,from the portfolio generation to analysis,and the efficiency of proposed model is verified.     

基于DSM的项目成组管理模式的研究

为克服传统的项目进度计划方法在基于多种约束条件的多项目管理环境下表现出的不足,引入设计结构矩阵(DSM)建立了基于DSM的项目成组管理模式,该模式能够将多项目自身信息、资源使用信息、各个作业之间的信息约束关系和资源约束关系等充分反映在一个矩阵中,从而便于进行多项目管理.用1个实例说明了该模型的实用性.     

带有交易成本的均值——方差——下半方差投资组合模型

投资组合选择研究为投资决策和风险管理提供了可量化的途径和科学决策的依据．本文引入了非凹非凸的典型交易成本函数,建立了含有交易成本函数的均值--方差--下半方差投资组合模型．考虑到不同的投资者对风险的厌恶程度不同,引入风险厌恶系数,把双目标的投资组合优化模型转化为单目标的投资组合优化模型,并运用教与学算法对模型进行了求解,得到不同收益下的最优投资组合,同时给出了投资组合的有效边界,最后对算法的优越性进行了分析,得到了比较好的仿真结果．     

A two-period model for selection of resilient multi-tier supply portfolio

A two-period decision-making model is developed for selection of resilient supply portfolio in a multi-tier supply chain under disruption risks. The planning horizon is divided into two aggregate periods: before and after the disruption. The resilience of the supply chain is achieved by selection ahead of time primary supply portfolio and by pre-positioning of risk mitigation inventory of parts at different tiers that will hedge against all disruption scenarios. Simultaneously, recovery and transshipment portfolios are determined for each disruption scenario and decisions on usage the pre-positioned inventory are made to minimise expected cost or maximise expected service level. Some properties of optimal solutions, derived from the proposed model provide additional managerial insights. The findings also indicate that the developed portfolio approach with an embedded network flow structure leads to computationally efficient stochastic mixed integer program with a strong LP relaxation.     

Forward and backward risk assessment throughout a system life cycle using dynamic Bayesian networks: A case in a petroleum refinery

In this paper, risk modeling was conducted based on the defined risk elements of a conceptual risk framework. This model allows for the estimation of a variety of risks, including human error probability, operational risk, financial risk, technological risk, commercial risk, health risk, and social and environmental risks. Bayesian network (BN) structure learning techniques were used to determine the relationships among the model variables. By solving a bi-objective optimization problem applying the genetic algorithm (GA) with the Pareto ranking approach, the network structure was learned. Then, risk modeling was performed for a petroleum refinery focusing on HydroDeSulfurization (HDS) technology throughout its life cycle. To extend the model horizontally and make it possible to evaluate the risk trend throughout the technology life cycle, we developed a dynamic Bayesian network (DBN) with three-time slices. A two-way forward and backward approach was used to analyze the model. The model validation was performed by applying the leave-one-out cross-validation method.