METAD: Monitoring Entites for Trend Analysis and Data mining

We propose to investigate methods for building METAD, a system for monitoring trends in order to identify entities that are suspicious or deviate from past or normal behavior. Our system will leverage our existing EntityBase technology, which provides the means to consolidate references to entities collected from disparate data sources. The system will first use trend analysis to find entities that deviate from normal or past behavior and then use data mining techniques such as group detection, clustering or guilt-by-association to refine and rank these entities by how suspicious they are. The system will be built with the analyst in mind and will be an interactive system that assists the analyst rather than being a black-box investigation tool.