rb.r

# Criando uma rede vazia
library(bnlearn)
dag = empty.graph(c("P", "N"))
class(dag)
dag

# Criando a estrutura da rede
arc.set = matrix(c("P", "N"), ncol = 2, byrow = TRUE, dimnames = list(NULL, c("de", "para")))
arc.set
arcs(dag) = arc.set
dag

P.lv = c("Ativo", "Reflexivo")
N.lv = c("Boa", "Ruim")

P.prob = array(c(0.5, 0.5), dim = 2, dimnames = list(P = P.lv))
N.prob = array(c(0.5, 0.5, 0.5, 0.5), dim = c(2,2), dimnames = list(N = N.lv, P = P.lv))

cpt = list(P = P.prob, N = N.prob)
cpt

bn = custom.fit(dag, cpt)
bn

# Dados utilizado como exemplo na dissertação
sample = rbn(bn, n = 10)
sample
class(sample)
sample = data.frame(P = c("Ativo", "Ativo", "Ativo", "Ativo", "Ativo", "Reflexivo", "Reflexivo", "Reflexivo", "Reflexivo", "Reflexivo"),
                    N = c("Ruim", "Ruim", "Ruim", "Ruim", "Boa", "Ruim", "Ruim", "Boa", "Ruim", "Boa"),
                    stringsAsFactors = TRUE)

bn = bn.fit(dag, sample)
bn

#   Bayesian network parameters
# 
#   Parameters of node P (multinomial distribution)
# 
# Conditional probability table:
#      Ativo Reflexivo 
#       0.5       0.5 
# 
#   Parameters of node N (multinomial distribution)
# 
# Conditional probability table:
#  
#       P
# N      Ativo Reflexivo
#   Boa    0.2       0.4
#   Ruim   0.8       0.6


cpquery(bn, event = (N == "Ruim"), evidence = (P == "Reflexivo" | P == "Ativo"))
# [1] 0.6942
cpquery(bn, event = (N == "Boa"), evidence = (P == "Reflexivo" | P == "Ativo"))
# [1] 0.3068

cpquery(bn, event = (P == "Ativo"), evidence = (N == "Boa"))
# [1] 0.3226247
cpquery(bn, event = (P == "Ativo"), evidence = (N == "Ruim"))
# [1] 0.5665432

cpquery(bn, event = (P == "Reflexivo"), evidence = (N == "Boa"))
# [1] 0.6615385
cpquery(bn, event = (P == "Reflexivo"), evidence = (N == "Ruim"))
# [1] 0.4313501