Unsupervised Learning : Estimating Outlier Scores Using Isolation Forest-Based Anomaly Detection Algorithms
This project explores the various isolation forest-based anomaly detection algorithms for estimating outlier scores. Methods applied in the analysis to identify abnormal points with patterns significantly deviating away from the remaining data included the Isolation Forest, Extended Isolation Forest, Isolation Forest with Split Selection Criterion, Fair-Cut Forest, Density Isolation Forest and Boxed Isolation Forest algorithms. Using an independent label indicating the valid and outlying points from the data, the different anomaly-detection algorithms were evaluated based on their capability to effectively discriminate both data categories using the area under the receiver operating characteristics curve (AUROC) metric.
