Implementation of Genetic Algorithm - Random Oversampling on the Random Forest Algorithm to Address Imbalanced Blood Donor Eligibility Data

Abstract

Purpose: This study aims to improve the accuracy of blood donor classification by addressing data imbalance using machine learning techniques. Accurate classification of donor eligibility is crucial for maintaining a reliable blood supply. To achieve this, the research explores the integration of the Random Forest algorithm with the Genetic Algorithm (GA) for feature selection and optimization, alongside Random Oversampling (RO) for data balancing.
Methods: The research employs the Random Forest algorithm combined with GA for feature selection and optimization. Additionally, Random Oversampling is applied to handle the class imbalance in the dataset. The model's performance is evaluated using 10-Fold Cross Validation. The dataset used in this study consists of blood donor records from the Indonesian Red Cross (PMI) in Samarinda City for 2023–2024.
Results: The application of Random Oversampling significantly improved the model’s accuracy, achieving 99.94%. However, the use of GA Feature Selection and GA Optimization independently did not result in notable improvements. Furthermore, when both techniques were applied together, the accuracy decreased to 98.78%.
Conclusions: The study confirms that Random Oversampling is highly effective in improving classification accuracy for blood donor eligibility. However, the integration of GA for feature selection and optimization did not yield additional benefits and even reduced accuracy when applied together. Future research could explore alternative feature selection and optimization methods to further enhance classification performance.