Detection of Class Imbalance Problem using Oversampling Technique

Now a day as the application area of technology increases the proportion of data and the nature also increases. One of such problem in data mining and machine learning techniques are class imbalance problem. Class imbalance problem is a problem where distribution of data largely belongs to one class while small or no data belongs to other class. Class imbalance problem is also known as skewed data problem. The data in real-world applications often has imbalanced class distribution where most of the classifier correctly classifies majority class data while they completely ignore minority. This is the problem associated with class imbalance problem. There many techniques used to solve class imbalance problem such data preprocessing, algorithmic approach and ensemble techniques. Data preprocessing gives better solution than other techniques. Data preprocessing techniques broadly classified into oversampling and under sampling technique. The disadvantage associated with under sampling is that it losses the information. So, mostly oversampling technique is used to balance the data. But disadvantage associated with the oversampling techniques is that it replicates unnecessary information. In this paper, proposed an approach to minimize the problem of replication of data associated with the oversampling technique.