Data mining for algorithmic asset management: an ensemble learning approach

Algorithmic asset management refers to the use of expert systems that enter trading orders without any user intervention. In particular, market-neutral systems aim at generating positive returns regardless of underlying market conditions. In this chapter we describe an incremental learning framework for algorithmic asset management based on support vector regression. The algorithm learns the fair price of the security under management by minimining a regularised ε-insensitive loss function in an on-line fashion, using the most recent market information acquired by means of streaming financial data. The difficult issue of learning in nonstationary environments is addressed by adopting an ensemble learning strategy, where a meta-algorithm strategically combines the opinion of a pool of experts. Experimental results based on nearly seven years of historical data for the iShare S&P 500 ETF demonstrate that satisfactory risk-adjusted returns can be achieved by the temporal data mining system after transaction costs.