您的位置：首页 > 论文页面
PCBN模型的参数估计及模拟抽样算法
发表时间：2018-04-13 浏览量：1581 下载量：195
| 全部作者： | 赵子然，王斌会 |
| 作者单位： | 暨南大学经济学院；暨南大学管理学院 |
| 摘 要： | 给出建立pair-copula贝叶斯网络（pair-copula Bayesian network，PCBN）模型时的参数估计和模拟抽样的步骤，并以具体的算法表示。参数估计和模拟抽样算法的关键在于利用h函数及其递推关系得到条件分布。研究发现，PCBN模型的参数估计及模拟抽样算法可在传统C-Vine模型参数估计及模拟抽样算法的基础上得出，其中模型的参数估计基于阶段极大似然法，模型的模拟抽样基于CPI Rosenblatt转换法（条件逆函数法），主要区别在于PCBN模型的算法循环遍历有向无环图（directed acycline graph，DAG）中的每个节点时需要判断排序在其之前的节点是否为其父节点。 |
| 关 键 词： | 数理统计学；算法；阶段极大似然法；模拟抽样；参数估计 |
| Title： | Algorithms of parameter estimation and sampling simulation in pair-copula Bayesian network model |
| Author： | ZHAO Ziran, WANG Binhui |
| Organization： | College of Economics, Jinan University; School of Management, Jinan University |
| Abstract： | This paper provides the steps of parameter estimation and sampling simulation in pair-copula Bayesian network (PCBN) model and represents them as algorithms. Trying to derive conditional distribution by h function and its recursive relations is the key to the algorithms of parameter estimation and sampling simulation. Based on the traditional C-Vine copula model, algorithms of parameter estimation and sampling simulation in PCBN model can be obtained easily. Algorithm of parameter estimation is based on sequential maximum likelihood method while algorithm of sampling simulation in PCBN model is based on CPI Rosenblatt conversion method (conditional inverse function method). Main difference between algorithms of PCBN and C-Vine is existed. When we loop through each node in directed acycline graph (DAG) of PCBN model, the parent node identification of the node before will be needed. |
| Key words： | mathematical statistics; algorithms; sequential maximum likelihood method; sampling simulation; parameter estimation |
| 发表期数： | 2018年4月第7期 |
| 引用格式： | 赵子然，王斌会. PCBN模型的参数估计及模拟抽样算法[J]. 中国科技论文在线精品论文，2018，11（7）：684-688. |
4
评论数 0请您登录
暂无评论
下载全文
多维论文