Database Reference
In-Depth Information
Section 7.1 has introduced a bank marketing dataset (
Figure 7.3
). This section
shows how to use the naïve Bayes classifier on this dataset to predict if the clients
would subscribe to a term deposit.
Building a naïve Bayes classifier requires knowing certain statistics, all calculated
from the training set. The first requirement is to collect the probabilities of all
class labels, . In the presented example, these would be the probability that
a client will subscribe to the term deposit and the probability the client will not.
From the data available in the training set,
and
.
The second thing the naïve Bayes classifier needs to know is the conditional
probabilities of each attribute given each class label , namely . The
training set contains several attributes:
job
,
marital
,
education
,
default
,
housing
,
loan
,
contact
, and
poutcome
. For each attribute and its possible
values, computing the conditional probabilities given or
is required. For example, relative to the
marital
attribute, the
following conditional probabilities are calculated.
After training the classifier and computing all the required statistics, the naïve
Bayes classifier can be tested over the testing set. For each record in the testing
set, the naïve Bayes classifier assigns the classifier
label
that maximizes
.
