Investing in emerging markets using neural networks and particle swarm optimisation

Emerging markets represent a particular challenge to both investors and those interested in developing automated trading strategies. However as well as exposing investors to potential risk, these markets can also offer high returns. Here, a stock trading model is developed for these markets, using both particle swarm optimisation and neural networks. Learning is in part driven by the Matthews correlation coefficient, a task-unspecific but effective fitness measure for unbalanced data sets, used by the authors in previous work, and in addition by a realistic measure of trading profit that incorporates transaction costs. The recommendations from the hybrid model are compared to those obtained from an industry standard stock selection method, with favourable results.