Imbalanced Data Classification using Sampling Techniques and XGBoost
Priyanka Lahoti, Ajeet Kumar · International Journal of Computer Applications · 2018
While implementing any machine learning algorithms it is good to have the descriptive knowledge of the dataset.In any dataset, in case having more than 90% of the data in target variable is from class 1 and the remaining data is from class 2. In such type of dataset, error evaluation metric accuracy is not going to help much.Having the unknown dataset with only class 1 itself gives more than 90% accuracy, which shows accuracy as evaluation metric should be ignored.Such a problem with highly skewed target outcome is known as an Imbalanced classification problem.There is a number of techniques to deal with imbalanced dataset.In this paper, we are interested to see how sampling techniques and XGBoost can be used while working with the Imbalanced dataset.