摘要：We extend Disjunctive Anomaly Detection (DAD) algorithm to handle various data distributions and models of cluster interactions. It enables efficient searching and explanation of multiple disease outbreaks occurring simultaneously. Detected clusters can span multiple values along multiple dimensions, and can impact any subset of dimensions in data. This type of search is known to be exponentially complex, so DAD uses approximations to enable fast processing of large data. We demonstrate DAD's ability to systematically outperform state-of-art alternatives including What's Strange About Recent Events (WSARE) and Large Average Submatrix (LAS) on data of scales and complexities typically encountered in biosurveillance applications.
