Grammatical Evolution-based ensembles for algorithmic trading

摘要: Abstract The literature on trading algorithms based Grammatical Evolution commonly presents solutions that rely static approaches. Given the prevalence of structural change in financial time series, implies rules might have to be updated at predefined intervals. We introduce an alternative solution ensemble models which are trained using a sliding window. structure combines flexibility required adapt changes with need control for excessive transaction costs associated over-trading. performance algorithm is benchmarked against five different comparable strategies include traditional approach, generation used single period and subsequently discarded, three alternatives ensembles voting schemes. experimental results, market data, show suggested approach offers very competitive results highlight importance containing costs.