3.8 Analyzing results
When running tests in WEKA, classifier output is generated [13]. Several values
and parameters are observable, below are the nine outputs analyzed in this report:
CCI - CCI stands for Correctly Classified Instances and is the foremost parameter
in which data has been compared in this report. CCI represents the percentage
of patients that have been correctly diagnosed, both with and without heart disease.
Kappa - Kappa measures the agreement of prediction with the true class. It
calculates the di!erence between the prediction with the observed agreement with
that expected by chance. The kappa statistic value is a value between 0-1. A value
greater than 0 means that the classifier is better than chance.
TP Average - Average of all patients who have been correctly diagnosed.
FP Average - Average of all patients who have been incorrectly diagnosed.
TP Healthy - True positive. The patient has been correctly diagnosed as
healthy (not having heart disease).
FP Healthy - False positive. The patient has been falsely diagnosed as
healthy (actually has heart disease).
TP Heart disease - True positive. The patient has been correctly diagnosed
with heart disease.
FP Heart disease - False positive. The patient has been falsely diagnosed
with heart disease (is actually healthy).
ROC Area - ROC Area is the area under the ROC curve. ROC curve shows
how well a classifier is at distinguishing between positive and negative instances.
ROC area can be used to evaluate the quality of the classifier and its ability to
separate positive and negative instances. ROC Area can be a value between 0.5
and 1, 1 being an optimal classifier and 0.5 being comparable with chance.