B. Characterizing parkinsonian gait: multi-class
classification task
Table III gives the prediction performance in the multiclass
problem of distinguishing among PD with significant
gait disturbance, PD with no significant gait disturbance, and
control. Because this classification task is no longer binary,
the performance metrics of sensitivity and specificity do not
apply. Precision is still applicable, and in this case, indicates
the rate at which data points classified as a member of a
certain class actually belong to that class. Recall is used to
measure prediction performance by indicating the rate at
which data points of a certain class can be correctly
identified as members of that class. In this classification task,
varying misclassification costs does not improve the
performance of the classifier in any metric without a
significant decrease in performance in other metrics.