Joint Research Conference
June 24-26, 2014
One of the key issues in monitoring supply chain operations is assuring a good detection capability, while maintaining a low rate of false alarms. Given massive amounts of data involved in such systems, it is important to focus on unfavorable conditions that are of practical, rather than statistical, importance. In conjunction with detection methodologies based on probabilistic data modeling and likelihood ratios, events of practical significance can be characterized in terms of targets and operating windows for key parameters of interest - and this is achieved via automated target-setting procedures. In this talk we discuss problems related to development and use of such procedures.