Abstract
Outlier detection or anomaly detection is an important and challenging issue in data mining, even so in the domain of energy data mining where data are often collected in large amounts but with little labeled information. This paper presents a couple of online outlier detection algorithms based on principal component analysis. Novel algorithmic treatments are introduced to build incremental PCA and kernel PCA algorithms with online learning abilities. Some preliminary experimental results obtained from a real-world household consumption dataset have produced some promising performance for the proposed algorithms.