Lagged correlation-based deep learning for directional trend change prediction in financial time series

作者：

Highlights：

• An approach to deep learning for trend prediction in noisy systems is proposed.

• Regression-based features are shown to add to the predictive accuracy of the model.

• Experiments on historical stock market data are validated and outperform baselines.

• Implications for modern financial economics and practitioners are discussed.

摘要

•An approach to deep learning for trend prediction in noisy systems is proposed.•Regression-based features are shown to add to the predictive accuracy of the model.•Experiments on historical stock market data are validated and outperform baselines.•Implications for modern financial economics and practitioners are discussed.

论文关键词：Lagged correlation,Deep learning,Trend analysis,Stock market

论文评审过程：Received 1 September 2018, Revised 25 October 2018, Accepted 20 November 2018, Available online 22 November 2018, Version of Record 27 November 2018.

论文官网地址：https://doi.org/10.1016/j.eswa.2018.11.027