Decision Forests - Data Mining with Decision Trees: Theory and Applications - page 135

simpler representations of the target attribute are induced. This manipula-
tion can be based on an aggregation of the original target's values (known
as Concept Aggregation ) or more complicated functions (known as Funct ion
Decomposition ).
Classical concept aggregation replaces the original target attribute with
a function, such that the domain of the new target attribute is smaller than
the original one.
Concept aggregation has been used to classify free text documents
into predefined topics [ Buntine (1996) ] . This application suggests breaking
the topics up into groups (co-topics) and then, instead of predicting the
document's topic directly, classifying the document into one of the co-topics.
Another model is then used to predict the actual topic in that co-topic.
A general concept aggregation algorithm called Error-Correcting Out-
put Coding (ECOC) which converts multi-class problems into multiple,
two-class problems has been suggested by Dietterich and Bakiri (1995).
A classifier is built for each possible binary partition of the classes.
Experiments show that ECOC improves the accuracy of neural networks
and decision trees on several multi-class problems from the UCI repository.
Theideatoconvert K class classification problems into K -two class
classification problems has been proposed by [ Anand et al . (1995) ] .Each
problem considers the discrimination of one class to the other classes. Lu
and Ito [ Lu and Ito (1999) ] extend Anand's method and propose a new
method for manipulating the data based on the class relations among the
training data. By using this method, they divide a K class classification
problem into a series of K ( K

1) / 2 two-class problems where each problem
considers the discrimination of one class to each one of the other classes.
The researchers used neural networks to examine this idea.
Function decomposition was originally developed in the Fifties and
Sixties for designing switching circuits. It was even used as an evaluation
mechanism for checker playing programs [Samuel (1967)]. This approach
was later improved by Biermann et al . (1982). Recently, the machine
learning community has adopted this approach. A manual decomposition
of the problem and an expert-assisted selection of examples to construct
rules for the concepts in the hierarchy was studied in [ Michie (1995) ] .
Compared to standard decision tree induction techniques, structured
induction exhibits about the same degree of classification accuracy with
the increased transparency and lower complexity of the developed models.
A general-purpose function decomposition approach for machine learning
was proposed in [ Zupan et al . (1998) ] . According to this approach, attributes

−