Standard classification algorithms are usually unable to correctly deal with imbalanced datasets because they are built under the premise of obtaining the maximum generalization ability. In this manner, these algorithms try to obtain general rules that cover as many examples as possible, benefiting the majority class, while more specific rules that cover the minority class are discarded because of its small presence in the whole dataset. In this way, the minority class examples are treated like noise and therefore, these samples are finally neglected in the classification