In this research, we look at the problem of detecting complex situations arising in Electronic Markets or e-markets. E-markets have been growing in size rapidly over the past few years. Large amounts of transactions are being generated from e-markets everyday so data analysis tools are required for several business processes such as market strategy evaluation or illegal trading activity detection. However, our literature review shows existing tools available today still cannot overcome all the main challenges such as dealing with a large amount of incoming real-time data from multiple market feeds and providing a user with no programming skills the ability to extract data efficiently. This thesis proposes to address this problem using the Event Processing concept. We model an e-market as a distributed event-driven system. Therefore, we can refer to e-market transactions as events and use an Event Processing System (EPS) as a data analysis tool. To implement our solution, we propose a new EPS architecture that allows the integration of several existing EPSs (called slave EPSs) under a unified domain-specific user interface and execution environment. Since different EPSs rely on different data models and event pattern models, our proposal also includes a unified e-market data model and event pattern framework for defining, composing and executing event patterns. Selected common event patterns used for financial market surveillance in several stock exchanges have been used to evaluate the proposed work. The proposed event pattern framework has proved that it has the capability of expressing event patterns of varying complexities. In terms of the proposed EPS architecture, a system prototype has been successfully developed using two sophisticated commercial systems, Coral8 and SMARTS, as slave EPSs. The experiments performed involve the execution of selected event patterns against real historical data from the Stock Exchange of Thailand (SET). Our solution is cost-efficient and provides a number of benefits that can be used in practice. The proposed data model and event pattern framework can be used during the requirement gathering phase. The proposed EPS architecture provides users a unified interface with no programming skills for application development and the ability to customise and execute event patterns for different existing EPSs. Moreover, it can be used to facilitate the suitable EPS selection to achieve a more efficient event pattern detection process.
| Identifer | oai:union.ndltd.org:ADTP/187547 |
| Date | January 2008 |
| Creators | Mangkorntong, Piyanath (Aim), Computer Science & Engineering, Faculty of Engineering, UNSW |
| Publisher | Awarded by:University of New South Wales. School of Computer Science & Engineering |
| Source Sets | Australiasian Digital Theses Program |
| Language | English |
| Detected Language | English |
| Rights | Copyright Piyanath (Aim) Mangkorntong, http://unsworks.unsw.edu.au/copyright |
Page generated in 0.0013 seconds