CORN algorithm and existing pattern matching–based approach. All these assump-
tions allow us to construct a portfolio using only similar appearances of historical
market prices, without considering other factors, either technical or fundamental.
8.5 Summary
This chapter proposed a novel “CORrelation-driven Nonparametric learning”
(CORN) strategy for online portfolio selection, which effectively exploits the sta-
tistical correlations hidden in stock markets, and benefits from the exploration of
powerful nonparametric learning techniques. The proposed CORN algorithm is sim-
ple in nature and easy to implement, and has parameters that are easy to set. It also
enjoys the universal consistency property. Our empirical studies on real markets, in
Part IV, show that CORN can substantially beat the market index and the best stock,
and also consistently surpasses a variety of state-of-the-art algorithms.
Currently, the proposed CORN can capture the linear relationship between two
market windows, and it is possible to further capture their nonlinear relationship.
Although high return strategies are often associated with high risk, it would be more
attractive to develop a strategy that can manage the risk properly without slashing
too much return. As an extension to this work, we are currently developing such risk-
limiting strategies for CORN. In future, we plan to investigate theoretical insights
of the algorithm and examine its extensions to improve the performance with high
transaction costs.
T&F Cat #K23731 — K23731_C008 — page 57 — 9/28/2015 — 21:18
..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.