Inferring Market Strategies: Applying Data-Mining to Analysis of Financial Markets

It has become increasingly common to model financial markets using frameworks which better capture their behavior than the excessively simplistic traditional frameworks. Key concepts in these new frameworks are evolution, complex systems and data mining, each with their associated characteristic analysis. In particular, data mining provides extremely useful tools for potentially extracting knowledge from the huge quantity of data available in financial markets. In this paper we present a new methodology for inferring, using market data, whether or not agents with similar performance are using similar trading strategies and by that to try to understand why certain agents are more successful than others. Put another way, we use data mining to look for "footprints", in the time series of price, that characterize the distinct trading strategies, and that are generated by their trading activity. One way to look at this is as a classification problem, where we try to classify agents with similar performance, determining if they are found in the same region of a discrete, multi-dimensional space composed of variables that are derived from the market data.