A Data Driven Machine Learning Approach to Discovering Rules of Price Behavior in a Financial Market Simulation

Abstract

The field of agent-based simulation of financial markets has grownconsiderably in the last decade. However, the interpretation ofsimulation results has received far less attention. Typically, theresults of a large number of simulations are reduced to one or twosummary statistics, such as sample moments. While suchsummarization is useful, it overlooks a vast amount of additionalinformation that might be gleaned by examining patterns ofbehavior that emerge at lower levels. In this paper we propose anapproach to interpreting simulation results that involves the use ofso-called data mining techniques to identify the rules of behaviorthat govern an underlying system. We demonstrate the approachby using data from a single run of an order market simulation toderive rules about the behavior of prices in that simulation.