Classification algorithms are meant to discover the underlying pattern of training data which can then be used to predict the outcome of unseen data. Hence, it is essential to have substantial amount of training data to build an efficient classification model. How-ever, in many discipline, real-world datasets are heavily skewed and some classes are significantly outnumbered by the other classes. In these situations, classification algo-rithms fail to achieve substantial efficacy while predicting these under-represented in-stances. To solve this problem, many variations of synthetic minority over-sampling methods (SMOTE) have been proposed to balance the dataset which deals with continu-ous features. However, for datasets with both nominal and continuous features, SMOTE-NC is the only SMOTE-based over-sampling technique to balance the data. In this paper, we present a new minority over-sampling method, SMOTE-ENC (SMOTE – En-coded Nominal and Continuous), in which, nominal features are encoded as numeric values and difference between two such numeric value reflects the amount of change of association with minority class. For minority class observations, experiments on 6 da-tasets show that, SMOTE-ENC method offers better prediction than SMOTE-NC. Also, this proposed method overcomes one of the major limitations of SMOTE-NC algorithm. SMOTE-NC can be applied only on mixed datasets (i.e., data with both continuous and nominal feature) and cannot function if all the features of the dataset are nominal. The new SMOTE-ENC algorithm has been generalized to be applied on both mixed dataset and on nominal only datasets.
Citation: SMOTE-ENC: A novel SMOTE-based method to generate syn-thetic data for nominal and continuous features
Dr. Matloob Khushi name@yourdomain.com (xxx) - xxxx