Automata Engine for Generating Investment Strategies

Abstract

This paper shows the process of generating financial automata via a novell automated system. Previous studies have argued that a generic financial automata generator would be difficult to develop. However, this paper documents extensions to previously attempted methods enabling the system to operate directly on financial primitives. The use of these primitives improve the capabilities and reusability of developed models without a significant increase in difficulity of implementation.

The above mentioned system is then tested by generating an investement strategy. This strategy is defined to exploit the momentum anomaly in the S&P-500 index. Furthermore, the exact same components of which this strategy is generated, are also used to generate a simple moving average indicator. This indicator is given to the strategy generation engine as the only additional component for generating the investment strategy.

To show the capabilities of the proposed system, no adjustments nor transformations are made to the input data. Instead the system is provided with components to perform this transformation on its own. This capability acompanied by lax restrictions to the search space for performant strategies and the sharing of components between different generation tasks has not been documented before in the financial context.