Comparative Analysis of Semi-Supervised Outlier Detection Algorithms

Outlier detection plays a crucial role in identifying anomalous data points within datasets. While unsupervised methods have been widely used for this task, situations may arise where a dataset is known to contain no outliers for a specific task. In such cases, semi-supervised outlier detection algorithms become imperative. This paper presents a comprehensive analysis of multiple semi-supervised outlier detection techniques. The study evaluates these algorithms across diverse datasets, considering variations in feature dimensionality, sample size, and outlier prevalence. As a main contribution, this study provides valuable insights into the strengths and limitations of semi-supervised outlier detection algorithms, aiding researchers and practitioners in selecting suitable methods for outlier detection tasks.