A Computational Approach to Predicting Economic Regimes in Automated Exchanges

We present a computational approach to identify dominant market conditions, such as over-supply or scarcity, and to predict market changes in automated exchange environments. Intelligent agents can learn the characteristics of prevailing economic conditions, or regimes, from historical data. Agents can then use real-time observable information to identify the current market regime and forecast upcoming market changes. We show that different market regimes can be effectively identified using our methodology. We also present preliminary work on a method to predict regime transitions. We experimentally validate our approach with data from the Trading Agent Competition for Supply Chain Management.