This is a simple network that relates
Gender to the characteristics of people. The characteristics represented in the network are
Eye Colour,
Hair Length,
Jewellery and
Height (see
Figure 1).
Figure 1: Gender Model
Open the Gender Model by clicking
here. You should now be in DBLi's network viewer. Notice that the probability distributions for all nodes within the Gender Model are
uniform. This is because the network is not yet populated with probabilities. One useful feature of Bayesian networks, however, is that they can learn probabilities from
data. For instance, if you surveyed a group of people and for each person recorded their
Gender,
Eye Colour,
Hair Length,
Jewellery and
Height, you could then use this data set to populate the Gender Model with probabilities. Let's have a look at how this is done.
Open the field data sets for the network by scrolling to the
Gender Model network in
View Networks and selecting the
icon to open a list of data sets for the Gender Model.
Figure 2: Gender Model Data Sets
There should be one data set listed called
Gender Model Data Set. Click on the
view data icon (
) to open this data set. The data set consists of five columns, one for each node in the Gender Model. Each row in the data set contains the data for a person. For instance, the first person in the data set is a
Female with
Medium hair length,
Green Blue eye colour, wears
Jewellery and falls within the height range
150 to 160 cm (see
Figure 3).
Figure 3: Gender Model Data Set
Now click on the
Update Network Probabilities button, located above the
Gender Model Data Set. A message should appear saying,
"The probabilities and utilities for the network have been updated". Close the View Field Data Sets window and return to the Gender Model. You will notice that the probabilities for the Gender Model have not changed. To update them, click on the
Refresh Network link in the network viewer. The probabilities in the network will update to reflect the data set (see
Figure 4).
Figure 4: Gender Model with updated probabilities
Now that the probabilities in the network have been updated, it can be used for
scenario analysis. Lets say we what to know the characteristics of
Males. To do this, select the state
Male for
Gender by clicking on it.. The network will update to show the characteristics for
Males (see
Figure 5). For the example shown in
Figure 5, there is a higher probability of the males included in the data set being
tall,
not wearing jewellery and having
short hair and
brown eyes. For
Females, the situation is different (see
Figure 6). They have a higher probability of being
shorter, a much higher probability of
wearing jewellery and higher probability of having
long hair.
Figure 5: Gender Model with Male scenario inserted
Figure 6: Gender Model with Female scenario inserted
