Pushing the limits of RFID: Empowering RFID-based Electronic Article Surveillance with Data Analytics Techniques

Abstract

False-positive classification is a central issue for RFID environments with limited process control, such as in-store settings. In the case of electronic article surveillance, false positives not only lead to incorrect inventory data but also trigger false alarms, which impair customer satisfaction. A typical counter measure is to reduce antenna power, which in turn leads to greatly diminished detection rates. In contrast, the present study investigates the applicability of data analytics to achieve high detection rates while retaining low false positives. In contrast to prior research, our test setting acknowledges the lack of process control in retail environments. We consider various walking paths and speeds as well as RFID tags concealed by shopping bags. To distinguish theft from non-theft events, we derive predictors, which are not just aggregations of the signal strength. Rather, individual reads are put into temporal relation to one another and are augmented with antenna information.