While Complex Event Processing (CEP) technology narrows down the search space to relevant events, it remains difficult to execute all these algorithms to screen all events at the same time. This is due in part to the growing complexity of investment portfolios and products, more mature risk management techniques and regulations, as well as the sheer size and speed of innovative news sources such as social media (e.g., Twitter).
To help overcome these challenges, we propose a rules-based approach to optimize news trading systems, where each event is screened by only a limited number of relevant text and data analysis algorithms. These are selected automatically from a set of executable components, and are integrated in iterative workflows as per a knowledge or rules database of market and event conditions. Events may also be chained together to detect emerging patterns through time, and provide performance benchmarks among comparable scenarios and strategies.
We begin the paper with a review of news trading and provide a typology of its various algorithms. We then address the challenge of integrating several data and text analysis algorithms, and discuss the factors making this task increasingly more complex. A more technical section is also included, briefly explaining the proposed system architecture and its value added for news trading. We close with a case study in the context of a merger arbitrage scenario, identifying the typical analysis workflows and how a rules-based integration of algorithms may provide more effective execution and fund management.