Classification is one of the fundamental tasks in data mining (Agrawal et al., 1993; Fayyad et al., 1996; and Han and Kamber, 2006), pattern recognition (Bezdek and
Pal, 2001; and Duda et al., 2001) and machine learning (Mitchell, 1997). Data mining
is generally defined as the process of extracting previously unknown and useful
knowledge from large databases. A data mining algorithm like classification rule mining
(Freitas, 2002; and Dehuri and Mall, 2006) needs to perform an exhaustive search to find a
set of rules satisfying the criteria such as accuracy, comprehensibility and interestingness.
In the classification task, each instance belongs to a certain pre-specified
class indicated by the value of the goal attribute. This attribute can take a number
of values; each of them corresponds to a class. Each instance consists of two parts
namely a set of predictor attribute values and a predicted attribute value. The former
are used to predict the value of the goal attribute.
The major drawbacks of these conventional classifiers are: (1) they classify
the data when there is some noise and uncertainties; and (2) they are unable to
handle multiple instances with overlapping attributes that belong to different classes.
To overcome the above limitations, fuzzy classifications are coming into picture.
Fuzzy classifiers can give a smooth classification boundary as compared to a
non-fuzzy classification method. Fuzzy classifiers based on fuzzy if-then rules are interpretable,
as the expert or user can easily verify the classification model by testing the
reasonability and reliability of the fuzzy rule sets in the classifiers. The real-life problems like
face and voice recognition, handwriting verification, etc., which are difficult to handle
by machines but are easily solved by human beings, can be solved using fuzzy
rule-based classifiers (Kuncheva, 2000). Hence, the fuzzy classifiers have become popular. |
