Innovative Outlier Detection with SEBOOST: A Novel Hybrid Ensemble Approach

Outlier detection is critical in numerous domains such as financial fraud detection, cybersecurity, and system surveillance. This study introduces SEBOOST, a novel hybrid ensemble algorithm that combines supervised and unsupervised machine learning techniques for improved outlier detection. The method leverages sequential ensembles to generate enriched feature spaces, applies feature selection using information gain to remove irrelevant features, and employs a supervised CATBoost classifier for final predictions. SEBOOST was rigorously tested on 21 real-world datasets, demonstrating its superior performance over state-of-the-art methods with higher precision, true positive rate, and AUC-ROC scores. These results indicate that SEBOOST as a robust and scalable framework for addressing outlier detection challenges across diverse applications.