A Tool for Comparing Outbreak Detection Algorithms

Abstract

Despite of the main objective of recent biosurveillance researches is bioterrorist attack threats, detection of natural outbreaks are also being tried to solve by governments all over the world. Such that, international foundations as WHO, OECD and EU publish public declaration about necessity of an international central surveillance system. Each data source and contagious disease carries its own patterns. Therefore, standardizing the process of outbreak detection cannot be applicable. Various methods have been analyzed and published on test results in biosurveillance researches. In general, these methods are the algorithms in literature of SPC and Machine Learning, although specific algorithms have been proposed like Early Aberration Reporting System (EARS) methods. Differences between published results show that, the characteristic of time series are tested with algorithm and the chosen parameters of this algorithm are also determine results. Our tool provides preprocessing of data; testing, analyzing and reporting on anomaly detection algorithms specialized at biosurveillance. These functionalities make it possible to use outputs for comparing algorithms and decision making.