Oversampling, Undersampling, Smote SVM dan Random Forest pada Klasifikasi Penerima Bidikmisi Sejawa Timur Tahun 2017

Abstract

Bidikmisi is tuition assistance from the government for high school graduates (SMA) or equivalent who have good academic potential but have economic limitations. Different from scholarships that focus on providing awards or financial support to those who excel. The achievement requirements for Bidikmisi are aimed at ensuring that Bidikmisi recipients are selected from those who truly have the potential and willingness to complete higher education. Given that the recipients of this bidikmisi must really be the right person, in this study a classification of the recipients of the 2017 bidikmisi in East Java will be carried out, in this study there is data that is not balanced the "Accepted" class is more than the "Not accepted" class. If the data is not balanced, almost all classification algorithms will produce much higher accuracy for the majority class than for the minority class. Researchers will handle class imbalances. The resampling technique used in research related to the prediction of bidikmisi recipients includes resampling techniques, namely Oversampling, Undersampling and SMOTE using two classification methods, namely SVM and Random Forest. The Oversampling technique was chosen because it does not reduce the amount of data but adds to the dataset that is lacking in the minority class. The Oversampling algorithm used is Synthetic Minority Over-sampling Technique (SMOTE), this algorithm was chosen from several resampling algorithms because SMOTE produces good accuracy and is effective in dealing with unbalanced classes because it reduces overfitting.