Data Stream Management System and Capital Market Surveillance

Abstract

Importance of real-time data analysis has been felt since early '90s and thus processing of streaming data (from either sensor networks or telecom switches or web and other disparate systems) is the demand of the industries worldwide. Quicker detection of fraudulent activities in a financial system is the order of the day. Thus capital market surveillance, if can be performed by using the streaming input of various trading transactions, without being stored, that would be beneficial to the regulatory authorities and stock exchanges. In this paper, we describe how stream processing using a data stream management system (DSMS) can be used for the above task and how effective would be that in terms of performance and latency. We present results obtained from using a commercial event stream processing system (IBM InfoSphere Streams platform) for certain typical fraud detection scenarios.