On-line Portfolio Selection via Mean Reversion Strategy

This paper presents a novel adaptive algorithm using mean reversion strategy without transaction cost. The antiocr algorithm only exploits the property of “reversal to the mean” and its performance not only significantly depends on the size of window but also fluctuates wildly according to the different size of window. To overcome these limitations, this proposed algorithm is designed to deal with the portfolio selection problem by fully exploiting both the price momentum and the price reversal in Chinese stock markets. Equipped with several parameters, the proposed mean reversion strategy can better track the changing stock market. Extensive experiments on real stock data from Chinese markets demonstrate the effectiveness of our strategies in comparison with the anticor algorithm without knowledge of the investment duration.