The Application of The Neighborhood Cleaning Rule in Conjunction with Random Forest, K-Fold Cross-Validation, and Grid Search for Addressing Imbalanced Datasets

Abstract

Finding a model that explains and separates data classes is the process of classification in data mining, which is used to guess the class of an item with an unknown class. Numerous strategies have been developed since categorization can be applied in a wide range of applications. But a common issue with classification is class imbalance. Data predictability suffers as a result of the issue of unbalanced classes. There are typically not an equal number of examples in each class in real-world categorization datasets. Class imbalance is not a problem when there are not significant differences in how the classes are distributed. Due to class imbalance, prediction models may skew in favor of the majority class, with the minority class contributing little to the model. One often used strategy for addressing class imbalance is the resampling technique. This study's objective is to put the Resampling Algorithm into practice. Neighborhood Cleaning Rule Random Forest K-Fold Tune Grid Search was carried out on a dataset that includes cases of Low Birth Weight Infants (BBLR) in Majene Regency and breast cancer diagnoses, which was posted on the UCI website. The Neighborhood Cleaning Rule (NCL), a data processing method, eliminates noise or other disturbances from datasets used for modeling or analysis. The F1-Score, G-Mean, Accuracy, and Sensitivity values from the model are good.