Handling Class Data Imbalance in Random Forest using the SMOTE and ADASYN Methods in Identify Economic Status

Abstrak

Imbalanced class data refers to an imbalance in the amount of training data between two different classes, where one class represents a large amount of data (majority class) while the other class represents a very small amount of data (minority class). Oversampling is a technique of balancing data by generating data in the minority class so that the amount is balanced with the data in the majority class with the aim of improving the classification results for the better. The oversampling method was chosen to avoid losing important information in the imbalanced dataset. Synthetic Minority Over-Sampling Technique (SMOTE) and Adaptive Synthetic Approach (ADASYN) are oversampling techniques that produce synthetic data and use the concept of adjacency in the algorithm. SMOTE and ADASYN can reduce the possibility of overfitting, which is a disadvantage of ordinary oversampling. These two methods will be combined with the Random Forest classification algorithm. This research was conducted to solve the problem of imbalance class data in the dataset of the Indonesian economic financial crisis. The data in this study uses 9 independent variables based on macroeconomic aspects and 1 dependent variable. The dependent variable used is categorized as binary, namely crisis and non-crisis conditions. The results of this study indicate that handling class imbalance data in the random forest (RF) classification algorithm results in better classification performance. ADASYN-RF produces the best performance with Accuracy, recall, precision, F1 Score, and ROC AUC scores of 98.26%, 66.67%, 72.22%, 65.57%, and 82.93%, respectively.

Identifier

• DOI: https://doi.org/10.5281/zenodo.17746930
• OpenAlex: https://openalex.org/W7107949133