A Density-Based Random Forest for Imbalanced Data Classification

Many machine learning problem domains, such as the detection of fraud, spam, outliers, and anomalies, tend to involve inherently imbalanced class distributions samples. However, most classification algorithms assume equivalent sample sizes for each class. Therefore, datasets pose a significant challenge in prediction modeling. Herein, we propose density-based random forest algorithm (DBRF) improve performance, especially minority classes. DBRF is designed recognize boundary samples difficult classify then use method augment them. Subsequently, two different classifiers were constructed model augmented original dataset dependently, final output was determined using bagging technique. A real-world material 33 open public used evaluate performance DBRF. On 34 datasets, could achieve improvements 2–15% over terms F1-measure G-mean. The experimental results proved ability solve classifying objects located on boundary, including classes, by taking into account density space.