Increasing the classification success in imbalanced data sets using artificial neural network

In data sets, imbalanced data sets emerge as a result of not having a balanced distribution among classes. One of the biggest problems encountered in these unbalanced data sets is classification success. While the classification success is close to high values in the majority class, inaccuracies and errors are observed in the classification success in the minority class. In this thesis, studies have been conducted to increase the success of classification in data sets with uneven distribution. Artificial neural networks have been used to increase the classification success. In this study, seven methods using artificial neural networks are proposed, and geometric mean and f measure metrics are used for classification results, and Friedman's means evaluation statistics measure is used to evaluate these metrics. In these methods, random samples were produced with artificial neural networks in the method with the most successful results, and these samples were limited to a threshold value. The results obtained from the methods in the thesis study were compared with the results of the original data set and the results of the basic SMOTE method. Successful results were obtained as a result of the comparison, and classification success was increased in unbalanced data sets.