Effectient Clustered Based Oversampling Approach to Solve Rare Class Problem

In data mining, problems are arises when we are applying some data mining techniques to real-life data, which frequently shows rare class problem. Another name of rare class is class imbalance. Class imbalance problem encounters when one class having more samples than other classes. . The minority class data are those data which occur less frequently. Most of the researcher used two techniques to tackle the issue of rare class. Those techniques are the preprocessing techniques which are named as under sampling and oversampling. In under sampling, Important data belongs to mjority has been eliminated. Oversampling is applied to solve rae class issue by replicating the sample, data belongs to minority. Rare class issue leads to misclassification of minority class sample. To overcome this issue, we proposed an efficient clustered based oversampling technique i.e. ECMO which will minimize the problem of mirroring of data associated with the oversampling technique and analyse the performance of ECMO technique.