The Effect of Imbalanced Data on Machine Learning Algorithms

Data imbalance is a widespread issue in data mining, affecting the reliability of machine learning algorithms. The connection between the degree of imbalance and algorithm performance needs to be better understood. To address this gap, our study examines how variations in imbalance levels influence the effectiveness of machine learning algorithms. Specifically, the goal of this study is to understand how class imbalance affects five popular machine learning algorithms—Random Forest, XGBoost, Support Vector Machine (SVM), Neural Network (MLP Classifier), and Deep Neural Network—depending on the various degrees of imbalance. The performance of algorithms is measured using F1 score, recall, and precision metrics. Our experiments revealed that XGBoost provides solid performance across the different degrees of imbalance, and it has the benefit of being the fastest among the five algorithms. Conversely, SVM struggled, particularly at lower thresholds.