Impact of Distance Measures on Imbalanced Classes for Rule Extraction

This paper identifies the supervised and unsupervised learning algorithms for imbalanced classes to extract rules for them. Most Machine learning algorithms work best when the number of instances of each class is roughly equal, but there are only specific algorithms to deal with the imbalanced classes. Imbalanced classification problems mean that the dependent (response) variable has an imbalanced proportion of classes. This results in biased predictions, misleading inaccuracies, and fluctuating performances in the datasets and henceforth, imbalanced classes need attention in machine learning. In the field of machine learning, the traditional model evaluation methods do not accurately measure the model performances when faced with imbalanced classes. However, in an imbalanced dataset, the minor class does not significantly contribute towards accuracy. Hence, accuracy should not be used to evaluate the models' performance for an imbalanced dataset. Thus, this paper discusses several data mining methodologies like Sampling, Clustering, Distance measures and Performance measures to extract more rules for the imbalanced classes.