AN ITERATIVE ESTIMATOR FOR PREDICTING THE HETEROGENEOUS ATTRIBUTE DATA SETS

The quality of the patterns which are the results of data mining is depends upon the quality of
data supplied to it. Most of the real time databases which are the sources for data mining posses the
deficiency in terms of completeness, correctness and consistency. Improving the quality of data in terms
of completeness is a challenging task. Many methods were proposed for imputing the missing values for
homogenous attributes. This paper proposes a mixed kernel function, which imputes the missing values
for the mixed attributes (the independent attributes are heterogeneous). The mixed kernel function is an
integrated unit which adopts the right method to impute the value for right attribute. For the categorical
attribute, our kernel function first assigns the mode value and the iteration continues till the right (most
probable) value gets converged and for the discrete attribute the mean value gets assigned and the
iteration continues till the most probable value is reached. The mixed kernel function is tested with a
sample database; it proves that it is performing well in terms of accuracy and iterations compared to
linear kernel function.