Outlier Ensembles

En savoir plus sur le livre

This book explores various methods for outlier ensembles, organizing them by the principles that enhance accuracy. It discusses techniques that improve the effectiveness of these methods and provides a formal classification, examining the conditions under which they excel. The authors analyze the theoretical and practical relationships between outlier ensembles and commonly used ensemble techniques in data mining, particularly in classification. They delve into the subtle differences between ensemble techniques for classification and outlier detection, highlighting how these nuances influence the design of algorithms for outlier detection. Designed for courses in data mining and related fields, the book includes numerous illustrative examples and exercises to aid classroom instruction. It assumes familiarity with the outlier detection problem and the general concept of ensemble analysis in classification, as many methods discussed are adaptations from classification techniques. Unique insights are offered through techniques like wagging, randomized feature weighting, and geometric subsampling, which are not found elsewhere. Additionally, the book analyzes the performance of various base detectors and their effectiveness. It serves as a valuable resource for researchers and practitioners aiming to optimize algorithmic design using ensemble methods.