Subspace Outlier Detection in High Dimensional Data using Ensemble of PCA-based Subspaces

Abstract

Outlier detection in high dimensional data faces the challenge of curse of dimensionality where irrelevant features may prevent detection of outliers. The Principal Component Analysis (PCA) is widely used for dimensionality reduction in high dimensional outlier detection problem. While no single subspace can to thoroughly capture the outlier data points; we propose to combine the result of multiple subspaces to deal with this situation. In this research, we propose a subspace outlier detection algorithm in high dimensional data using an ensemble of PCA-based subspaces (SODEP) method. Three relevant subspaces are selected using PCA features to discover different types of outliers and subsequently, compute outlier scores in the projected subspaces. The experimental results show that our ensemble-based outlier selection is a promising method in high dimensional data and has better efficiency than other compared methods.