CSV data format

Training (parameter learning) requires a dataset in CSV format. This CSV file must be formatted as follows:

• The first row of the CSV must be a header consisting of variable names

• The subsequent rows of the CSV must be data points that match the categories of the variables

Additionally, the CSV data must conform to some basic rules to pass validation.

For example, in the Sprinkler data set there are four possible variables: cloudy, rain, sprinkler, and wetGrass. Each of these variables consists of the following categories:

• cloudy: yes, no

• rain: yes, no

• sprinkler: on, off

• wetGrass: yes, no

An example CSV for this dataset would be (as a table):

cloudy	rain	sprinkler	wetGrass
no	no	off	no
yes	no	off	no
yes	yes	off	yes
yes	yes	off	yes
yes	yes	off	yes

In CSV format, this dataset would have the following form:

cloudy,rain,sprinkler,wet_grass
no,no,off,no
yes,no,off,no
yes,yes,off,yes
yes,yes,off,yes
yes,yes,off,yes

In the case of parameter learning, the variable names in the columns must match the variable names of the nodes in the graph. This is because parameter learning already assumes a model exists and updates probabilities accordingly.