Intelligent agents for adaptive security market surveillance.

Kun Chen,Xin Li,Baoxun Xu,Jiaqi Yan,Huaiqing Wang

DOI: https://doi.org/10.1080/17517575.2015.1075593

2017-01-01

Abstract:Market surveillance systems have increasingly gained in usage for monitoring trading activities in stock markets to maintain market integrity. Existing systems primarily focus on the numerical analysis of market activity data and generally ignore textual information. To fulfil the requirements of information-based surveillance, a multi-agent-based architecture that uses agent intercommunication and incremental learning mechanisms is proposed to provide a flexible and adaptive inspection process. A prototype system is implemented using the techniques of text mining and rule-based reasoning, among others. Based on experiments in the scalping surveillance scenario, the system can identify target information evidence up to 87.50% of the time and automatically identify 70.59% of cases depending on the constraints on the available information sources. The results of this study indicate that the proposed information surveillance system is effective. This study thus contributes to the market surveillance literature and has significant practical implications.