B. Evaluation metrics for imbalanced data sets
Instead of using the overall classification accuracy as a
single evaluation criterion, we use a set of assessment metrics
related to receiver operating characteristics (ROC) graphs [31]
to evaluate the performance of ADASYN algorithm. We use
ROC based evaluation metrics because under the imbalanced
learning condition, traditional overall classification accuracy
may not be able to provide a comprehensive assessment
of the observed learning algorithm [17] [31] [32] [33] [6]
[34] [16]. Let {p, n} be the positive and negative testing
examples and {Y,N} be the classification results given by
a learning algorithm for positive and negative predictions. A
representation of classification performance can be formulated
by a confusion matrix (contingency table) as illustrated in
Fig. 2. We followed the suggestions of [15] [34] and use the
minority class as the positive class and majority class as the
negative class.