Performance of Support Vector Machine in Imbalanced Data Set

This paper considers the problem of an imbalanced data set, in which one class has a significantly larger number of samples than the other. Since it is necessary to determine as accurately as possible the classification accuracy of a class that has a small number of samples, different methods are used in the paper to measure the performance of the classifier. We used real data set obtained from the transaction through a payment card, in which a certain number of transactions was generated by unauthorized persons. SVM algorithm was used for classification, with the best operating parameters determined during cross-validation. The paper presents the results obtained for measuring the performance of SVM algorithm.