UNDERSTANDING THE INFLUENCE OF OUTLIERS ON MACHINE LEARNING MODEL INTERPRETABILITY

Abstract

Outliers are data points that deviate significantly from the majority of a dataset and can profoundly impact the performance and interpretability of machine learning models. This study explores the influence of outliers on model interpretability, highlighting how they can skew learning patterns, distort feature importance, and introduce bias in model explanations. This research underscores the importance of effective outlier detection and management strategies by examining various types of outliers and their effects on different models. The study also reviews existing methods for identifying and handling outliers, including statistical techniques, machine learning approaches, and deep learning models, to enhance the reliability and transparency of machine learning systems. The findings emphasize the need for a balanced approach that considers both model accuracy and interpretability, ensuring that machine learning models remain trustworthy and informative in the presence of outliers.