GOS: A Genetic OverSampling Algorithm for Classification of Quranic Verses

Abstract

Imbalanced classes problem is a problem in many datasets in real applications, where one class "minority class" contain few numbers of samples and the other "majority class" contain many numbers of samples. It is difficult to build a training model to classify the imbalanced classes correctly due to tending the accuracy of classification to the majority class. In this paper, a new technique is called "GOS: a Genetic OverSampling algorithm", is proposed using a genetic algorithm. A genetic algorithm is applied to oversample the imbalanced datasets and to improve the performance of imbalanced classification. This improvement is achieved due to adjusting the locations of samples in the minority class in the optimal places. According to the experimental results obtained, the GOS algorithm outperformed other techniques used widely in the imbalanced classification field.