Outlier Detection in Online Gambling

Abstract

Data mining is field that is increasing in importance and width of application day by day. A sub-domain of data mining, the anomaly detection is also rising in importance the last years. Although discovered a long time ago, the last 5 or 10 years the uses of anomaly detection are increasing, therefore making it a useful technique to discover fraud, network intrusions, medicine side effects and many other useful anomalies within a wide set of data. The task of this master thesis is to find a more optimal anomaly detection technique to uncover fraudulent use or addictive playing in the transaction data of online gambling websites. This work is conducted on behalf of a Swedish company that is occupied in the field of data mining. For the needs of this work an anomaly detection method has been adapted, implemented and tested. The evaluation of this method is done by comparing the results it brings with the anomaly detection technique currently used for the same purpose.