BNLearn：如何将高斯贝叶斯网络的估计参数与其条件结构合并？

Question

我使用 iamb 函数定义了高斯贝叶斯网络的结构，然后使用 bn.fit.

估计了节点的系数

图书馆

library(bnlearn)

数据

{  C       E       G       N       V       W
48.83   51.48   42.64   54.1    42.96   41.96
48.85   73.43   40.97   60.07   65.29   48.96
67.01   71.1    52.52   51.64   63.22   62.03
37.83   49.33   56.15   49.01   47.75   38.77
55.3    49.27   63.55   54.62   60.57   56.66
56.12   48.72   66.02   43.95   55.54   52.39}

代码

# Definition of mandatory and forbidden nodes - here the white list
wl = data.frame(from = c("E","G","V","W","N"), to = c("V", "V","W","C","C"))

# Definition of the constrained network
network <- iamb(Data, test = "cor", whitelist = wl)

# Estimation of the coefficients according to the structure of the network
est.para <- bn.fit(network, data = Data)

问题是est.para是一个列表而不是可以绘制的GBN等等。我想知道如何合并网络和估计参数？

Answer 1

如果你想要一些网络图来显示除连接之外的一些额外信息，你可以使用 strength.plot。按照你的例子：

library(Rgraphviz)

strength <- arc.strength(network, Data)
strength.plot(network, strength, shape = "ellipse")

如果绝对需要使用 GBN est.para 参数的结果，您可以使用 graphviz.plot 参数来突出边和节点（可以使用 edgeRenderInfo 和 nodeRenderInfo)。举个例子，您可以使用参数来选择边缘的宽度：

library(data.table) 

plot <- graphviz.plot(network, shape = "ellipse")

arc.sizes <- data.table(network$arcs)
arc.sizes[, edge.name := paste0(arc.sizes$from, "~", arc.sizes$to)]
arc.sizes[, param := abs(est.para[[to]]$coefficients[[from]]), by = .(from, to)]
arc.sizes[, lwd := 5*((param - min(param))/(max(param) - min(param)))]

lwd <- as.vector(arc.sizes$lwd)
names(lwd) <- arc.sizes$edge.name
edgeRenderInfo(plot) <- list(lwd = lwd)

renderGraph(plot)

BNLearn：如何将高斯贝叶斯网络的估计参数与其条件结构合并？

BNLearn: How to merge the estimating parameters of a Gaussian Bayesian network with its conditional structure?

r

bayesian

bayesian-networks

bnlearn

图书馆

数据

代码