Semi Supervised Under-sampling: a Solution to the Class Imbalance Problem for Classification and Feature Selection

Most medical datasets are not balanced in their class labels. Furthermore, in some cases it has been noticed that the given class labels do not accurately represent characteristics of the data record. Most existing classification methods tend not to perform well on minority class examples when the dataset is extremely imbalanced. This is because they aim to optimize the overall accuracy without considering the relative distribution of each class. The class imbalance problem can also affect the feature selection process. In this paper we propose a cluster based under-sampling technique that solves the class imbalance problem for our cardiovascular data. Data prepared using this technique shows significant better performance than existing methods. A feature selection framework for unbalanced data is also proposed in this paper. The research found that ReliefF can be used to select fewer attributes, with no degradation of subsequent classifier performance, for the data balanced by the proposed under-sampling method.