This work can be extended in several ways. First, if there exists
distances between concept labels in the hierarchical tree, further
work should be undertaken to design a decision tree that not only
maximizes the accuracy and precision, but also minimizes the distance between the real label and predicted label. Second, if the data
with hierarchical labels could be extended as a directed acyclic
graph (DAG), we could explore building decision trees that classify
data with more complex analyses. In DAG structures, a child node
may have more than one parent; hence, the class labels in deeper
levels may be induced from more instances than their ancestors. In
all the above extensions, as in this paper, the goal is to find the best
tradeoff between precision and accuracy.