Classification is a model for making predictions based on existing data. Unbalanced data leads to misclassification or modeling errors where the data is irrelevant and results in poor classification modeling. The poor classification model is caused by an imbalance in the data on the classification label, so it is necessary to balance the data as a solution to overcome this problem. The methods used to deal with data imbalance are Random Undersampling and MWMOTE. The aim is to see the implementation of Random Undersampling and MWMOTE work well in dealing with unbalanced datasets and to know the performance and accuracy in modeling. The dataset used is an open source dataset from Kaggle which consists of Diabetes data, Bank Turnover data, Stroke data, and Credit Card data with various data ratios, with the aim of overcoming the problem of data imbalance. Model evaluation was carried out using confusion matrix and decision tree algorithms by looking at the values of precision, recall, f-measure, and accuracy of the original data, the Random Undersampling method, and MWMOTE. In the original data with 48.86% precision, 54.90% recall, 51.73% f-measure, and 85.30% accuracy. Random Undersampling can overcome data imbalance problems with 76.28% precision, 76.74% recall, 76.48% f-measure, and 76.21% accuracy. MWMOTE can solve data imbalance problems with 86.04% precision, 87.30% recall, 86.66% f-measure, and 86.61% accuracy. It can be concluded that the MWMOTE method is better than the Random Undersampling method because the evaluation average of the Confusion Matrix Random Undersampling method is smaller than the MWMOTE method.