In bayes-modeling/wrmbn: Water Resource Management using Bayesian Network

Preprocess data

library(wrmbn)

data("data")
data("structure")

uncorelate_node <- c("MTL", "QMX", "HCL", "CMB", "CTL", "CDX", "CBT")
for(node in uncorelate_node) {
  data <- data[, -which(colnames(data) == node)]
}

for(node in uncorelate_node) {
  if(length(which(structure$from == node)) > 0) {
    structure <- structure[-which(structure$from == node), ]
  } else if(length(which(structure$to == node)) > 0) {
    structure <- structure[-which(structure$to == node), ]
  }
}

type <- "continuous"
time_column <- "date"
continuous_variable_names <- setdiff(colnames(data), time_column)
discrete_variable_names <- c()
desire_layers <- 3
normalize_type <- "mean_normalization"
preprocessed <- preprocess_timeseries_data(data, type, time_column,
                                       continuous_variable_names, discrete_variable_names, desire_layers,
                                       normalize_type, quantile_number = -1, na_omit = TRUE)

Structure Filtering

library(wrmbn)

continuous_data <- preprocessed$data
desire_layers <- preprocessed$desire_layers
quantile_number <- preprocessed$quantile_number
continuous_variables <- preprocessed$continuous_variables
discrete_variables <- preprocessed$discrete_variables
known_structure <- structure
is_variable_only <- TRUE
corr_threshold <- 0
is_blacklist_internal <- TRUE
is_blacklist_other <- TRUE
custom_blacklist <- NULL
custom_whitelist <- NULL

bl_wl <- wrmbn::get_continuous_structure_filter(continuous_data, desire_layers, quantile_number,
                                            continuous_variables, discrete_variables,
                                            known_structure, corr_threshold, is_blacklist_internal,
                                            is_variable_only, is_blacklist_other,
                                            custom_blacklist, custom_whitelist)

Select alogrithms to learning

library(wrmbn)
data <- preprocessed$data
#algorithms <- c("gs", "hc", "tabu", "iamb", "inter.iamb", "fast.iamb")
algorithms <- c("hc", "tabu")
target <- "MBT_3"
wl <- bl_wl$whitelist
bl <- bl_wl$blacklist

result <- wrmbn::cross_validation_learning_algorithms(data, algorithms, wl, bl, target, n_cluster = 4, debug = TRUE)
#boxplot(result$gs, result$hc, result$tabu, result$iamb, result$inter.iamb, result$fast.iamb, names = algorithms)
boxplot(result$hc, result$tabu, names = algorithms)

Learning Model

library(wrmbn)
training_type <- preprocessed$type
data <- preprocessed$data
number_layers <- preprocessed$desire_layers
bl <- bl_wl$blacklist
wl <- bl_wl$whitelist
n_cluster <- 4
#algorithms <- c("gs", "hc", "tabu", "iamb", "inter.iamb", "fast.iamb")
algorithms <- c("hc", "tabu")
trained_models <- wrmbn::training_model(training_type, data, number_layers, bl, wl, n_cluster, algorithms, number_bootstrap = 500, debug = FALSE)