一种带隐变量贝叶斯网的事件诱因估计方法  

Method of Event Inducement Estimation Based on Bayesian Network with a Latent Variable

作　　者：何勇 吴鑫然 岳昆[1] HE Yong;WU Xin-ran;YUE Kun(School of Information Science and Engineering,Yunnan University,Kunming 650500,China)

机构地区：[1]云南大学信息学院,昆明650500

出　　处：《小型微型计算机系统》2022年第8期1609-1614,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金-云南联合基金重点支持项目(U1802271)资助;云南省万人计划"青年拔尖人才"计划项目(C6193032)资助;云南省教育厅科学研究基金项目(2020Y0010)资助.

摘　　要：事件诱因是诱导事件发生的因素,从事件特征数据构建事件诱因模型,进行事件诱因估计,是解决舆论控制、精准决策支持和用户行为定向等问题的重要基础.本文以公共突发事件为背景,以贝叶斯网为不确定性知识表示和推理的基本框架,以多值隐变量来描述事件诱因的多个取值,提出一种基于带隐变量贝叶斯网(隐变量模型)的事件诱因模型构建方法,进而利用概率推理算法估计事件诱因.针对事件诱因存在多个取值的问题,本文基于分支限界思想提出最优取值子集提取算法.建立在真实数据集上的实验结果表明,本文提出的事件诱因模型构建方法及相应的诱因估计方法是有效的.Modeling and estimating event inducements inducing events from feature data is significant for public opinion control,precise decision support,and user behavior orientation,etc.In this paper,we adopt Bayesian Network as the framework for the representation and inference of knowledge with uncertainty under the background of public emergencies.In particular,we use a multi-valued latent variable to describe the multiple values of the event inducement.Then,we propose a method for constructing an event inducement model based on Bayesian Network with a latent variable(a.k.a.latent variable model).Following,we estimate the event inducements by probabilistic inferences over the constructed model.Aiming at the multiple possible values of the event inducement,we give a branch and bound algorithm to extract the optimal subset of inducement.Experimental results on real datasets show the effectiveness of our proposed method.

关 键 词：事件诱因 贝叶斯网 隐变量 诱因估计 概率推理 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]