|
| | | |
Unsupervised Fraud Detection in Medicare Australia
Tang, M., Sumudu, B., Mendis, U., Murray, D. W., Hu, Y and Sutinen, A.
Fraud detection is a fundamental data mining task with a wide range of practical applications. Finding rare and evolving fraudulent claimant behaviour in millions of electronic Medicare records poses unique challenges due to the unsupervised nature of the problem. In this paper, we investigate the problem of efficiently and effectively identifying potential non-compliant Medicare claimants in Australia. We propose an unsupervised and data-driven fraud detection system called UNISIM. It integrates various techniques, such as feature selection, clustering, pattern recognition and outlier detection. By utilising the beneficial properties of these techniques, we are able to automate the detection process. Additionally, useful temporal patterns are extracted from the existing data for future analysis. Through extensive empirical studies, UNISIM is shown to effectively identify suspects with highly irregular patterns. Additionally, it is capable of detecting groups of outliers. . |
Cite as: Tang, M., Sumudu, B., Mendis, U., Murray, D. W., Hu, Y and Sutinen, A. (2011). Unsupervised Fraud Detection in Medicare Australia. In Proc. Australasian Data Mining Conference (AusDM 11) Ballarat, Australia. CRPIT, 121. Vamplew, P., Stranieri, A., Ong, K.-L., Christen, P. and Kennedy, P. J. Eds., ACS. 103-110 |
(from crpit.com)
(local if available)
|
|