Link Analysis Technologies for Differing Data Sources

AI Technologies for Homeland Security

Volume

Papers from the 2005 AAAI Spring Symposium

Contents

Analysts supporting homeland security via threat network analysis have a wide range of data sets at their disposal, and many link analysis technologies are capable of generating useful analytical information. While each data set provides critical information, and each tool has a unique set of strengths, not all tools are applicable to all data sets, even if the data is formatted in a way that the tool can use. Link analysis technologies do not perform equally well on all data set types, yet we have only begun to scratch the surface of exploring how the various technologies available perform against the varying data types, and the most appropriate way to measure the effectiveness of these technologies. Further, we must understand the implications of combining these data sets, with and without noise models for the various data sets. It remains to be determined whether data sets should be merged before analysis—the current trend in the face of increased data sharing among agencies—or if better results emerge from combining results derived from the individual data sources independently.