Handling minority class instances using classification technique

Real time applications deal with huge and rapidly changing data. It is difficult to extract knowledge from such huge and rapidly changing data. The problem arises when the focus is on examples with less number of observations. This is nothing but data imbalanced problem. The imbalanced learning focuses on data with very less number of observations. So to correctly classify the data with such less number of observations is a challenge, as classifiers built on such imbalanced data may tend to misclassify the minority class instances. Classification of data with such inherent complex characteristics requires iterative learning module. So best classifier needs to be selected for classification. This paper provides an overview of various approaches for handling minority class data and preliminary work related to the system which would eliminate the irrelevant attributes and accurately classify minority instances.