作者: Vince Vella , Wing Lon Ng
DOI: 10.1109/CIFER.2014.6924110
关键词:
摘要: We extend Neural Network (NN) trading models with an innovative and efficient volatility filter based on fuzzy c-means clustering algorithm, where the choice for number of clusters, a frequent problem cluster analysis, is selected by optimizing global risk-return performance measure. Our algorithm automatically extracts rules from past trades taking into account predicted return size intraday time varying realized volatility, latter used as proxy uncertainty. The model identifies unique scenarios subsequently creates dynamic visually apprehensible search space to control algorithmic decisions. results show that this method can be successfully applied support high-frequency strategies, outperforming both standard NN buy-and-hold models.