Online Portfolio Selection

Online portfolio selection is a dynamic asset allocation method based on online learning techniques. The portfolio selection task is transformed into a sequencial decision problem. And different approaches would be adopted to achieve the goal. In this paper, we focus on one of the ”patternmatching” based approaches, correlation-driven nonparametric learning. We improved the exsiting strategy and built up our own Correlation-driven Nonparametric Decay(CORN-D) model. We applied our algorithms to both stock dataset and index dataset and conduct strategy analysis based on MSCI index dataset. And finally we applied the strategy to ETF dataset for backtesting with the chosen parameters and selected methods. We finally drew the conclusion that our strategy is suitable for trading portfolios such as ETFs instead of trading single stocks and performs well under high volatility periods. We also found out that our strategy performed extremely well when there are no transaction costs but the performance became worse when transaction costs included.