Abstract:
Classification analysis on data that contains imbalanced data must first begin with a
process of balancing data classes to avoid misclassification. This study applies the
Synthetic Minority Oversampling Technique (SMOTE) method in overcoming
imbalanced data in cases of classification analysis of household welfare status with the
poor category as the minority class and non-poor as the majority class. The classification
method used is Binary Logistic Regression. This study uses 18 variables from the
household welfare status data and the variables that influence it. Based on the results of
the analysis it is known that the SMOTE method provides good performance in
overcoming imbalanced data for the classification of household welfare status. The
results for accuracy, sensitivity, specificity, G-mean, and AUC in binary logistic
regression were 84.28%, 66.67%, 84.61%, 0.75 and 0.84.