Using grocery sales data for the detection of bio-terrorist attacks

In this paper we explore the potential of using sales of grocery items data for early detection of epidemiological outbreaks and bio-terrorism attacks. These data are of special importance, as they are illustrative of non-symptom specific data that are expected to arrive earlier than medical data commonly used for such purposes. We explore the characteristics of such data and create a detection algorithm that detects irregular patterns of purchases that indicate an epidemiological outbreak. We show that it is feasible to use non-specific syndrome data, such as over-the-counter medication sales, for early detection of bio-terrorism attacks. Our conclusions are based on experiments with a theoretical simulation of a large anthrax outbreak. The proposed detection system consists of several layers and combines methods from signal processing, machine learning, statistics, and quality control. Finally,