國立成功大學 統計與資料科學學系

Abstract

This study develops methods to detect anomalous transactions linked with fraud in food stamp purchases through order statistics methods.  The methods detect clusters in the order statistics of the transaction amounts that merit further investigation.  Our techniques use scan statistics to determine when an excessive number of transactions occur (cluster), which is historically linked to fraud.  A scoring paradigm is constructed that ranks the degree in which detected clusters and transactions within clusters are anomalous among approximately 250 million total transactions.