Integration of Domain Knowledge for Outlier Detection in High Dimensional Space

摘要

The role of outlier or anomaly detection is to discover unusual and rare patterns in data. However, while the notion of "rare" can be quantified, what is "unusual" is high subjective and domain dependent. The focus of the proposal is to determine whether logic models like Probabilistic Relational Model, Default Reasoning and Ontologies can be used to integrate domain knowledge in the outlier discovery process.