sdglinkage (version 0.1.0)

gen_bn_learn: Generate synthetic data using BN learning.

Description

gen_bn_learn uses Bayesian structure learning to simultaneously learn the dependencies and the value of the parameters from the input data.

Usage

gen_bn_learn(training_set, structure_learning_algorithm, evidences = NA)

Arguments

training_set

A data frame of the training data. The generated data will have the same size as the training_set.

structure_learning_algorithm

A string of the structure learning algorithm from bnlearn.

evidences

A string of evidence that is used to constraint the sampling of the generated data.

Value

The output is a list of three objects: i) structure: the structure of the learned BN indicating the relationship between the variables (a bn-class object); ii) fit_model: the fitted model showing the parameter distributions between the variables ((a bn.fit) object and iii) gen_data: the generated synthetic data - if there is evidence to constraint the values for some of the variables, the generated synthetic data will be sampled accroding to the criteria.

Details

The structure learning algorithms including: 'tabu' for Tabu search, 'hc' for hill-climbing, 'pc.stable' for PC, 'gs' for Grow-Shrink, 'iamb' for Incremental Association, 'fast.iamb' for Fast Incremental Association, 'inter.iamb' for Interleaved Incremental Association, mmhc' for Max-Min Hill-Climbing, 'rsmax2' for Restricted Maximization, 'mmpc' for Max-Min Parents and Children, 'si.hiton.pc' for Semi-Interleaved HITON-PC, chow.liu' for Chow-Liu and 'aracne' for An Algorithm for the Reconstruction of Gene Regulatory Networks in a Mammalian Cellular Context.

Examples

# NOT RUN {
adult_data <- split_data(adult[1:100,], 70)
bn_learn1 <- gen_bn_learn(adult_data$training_set, "hc")
bn_evidence <- "age >=18 & capital_gain>=0 & capital_loss >=0 &
               hours_per_week>=0 & hours_per_week<=100"
bn_learn2 <- gen_bn_learn(adult_data$training_set, "hc", bn_evidence)

# }