BuildBayesNet {rWMBAT} | R Documentation |
BuildBayesNet selects features and metafeatures based on mutual info
BuildBayesNet(data, class, ffactor, drop)
data |
integer array containing the data used to build the Bayes net,cases in rows, variables in cols |
class |
double column vector, the known class variable for each case |
ffactor |
multiple of auto MI to use to threshold C->V connections |
drop |
double, MI loss percentage threshold for testing independence. Set to.75 and adjust to filter too few/too many variable-to-variable connections |
This function takes a set of training data and an additional variable called "class" and tries to learn a Bayesian Network Structure by examining Mutual Information.
a matrix of zeros and ones, where one in row i, column j denotes a directed link in a Bayesian network between variable i and variable j. The class variable is the last row/column.
CALLED FUNCTIONS: automi(finds an MI threshold based on data) findmutualinfos(finds all values MI(VC), MI(VV) and MI(VC|V))
Karl Kuschner, Qian Si and William Cooke , College of William and Mary, Dept. of Physics, 2009.
http://kwkusc.people.wm.edu/dissertation/dissertation.htm
data(traingrpbin, traingrpclass, ff, drop) #load example data in package adjmat <- BuildBayesNet( traingrpbin, traingrpclass, ff, drop )