Modelling financial time series using Grammatical Swarm

In this paper we employ a methodology based on Grammatical Swarm (GS) in producing models of financial returns. The models produced are used in trading single stocks in high frequency. The performance of the models produced using GS is compared to the performance of models produced using Grammatical evolution (GE) and models produced using GS outperform models produced using GE. Further analysis shows that the models produced using GS are better than a random strategy, and an AR model picked using Aikake Information Criteria (AIC).