Aggregating expert advice strategy for online portfolio selection with side information

Abstract

Online portfolio selection is an important fundamental problem in computational finance, which has been further developed in recent years. As the financial market changes rapidly, investors need to dynamically adjust asset positions according to various financial market information. However, existing online portfolio strategies are always designed without considering this information, which limits their practicability to some extent. To overcome this limitation, this paper exploits the available side information and presents a novel online portfolio strategy named “WAACS”. Specifically, all the constant rebalanced portfolio strategies are considered as experts and the weak aggregating algorithm is applied to aggregate all the expert advice according to their previous cumulative returns under the same side information state as the current period. Furthermore, WAACS is theoretically proved to be a universal portfolio, i.e., its growth rate is asymptotically the same as that of the best state constant rebalanced portfolio, which is a benchmark strategy considering side information. Numerical experiments show that WAACS achieves significant performance and demonstrate that considering side information improves the performance of the proposed strategy.