Effects of the validation set on stock returns forecasting

作者：

Highlights：

• A multiple combination of parameters is better than a single one for deep learning.

• Specific validation sets create better models than diverse sets.

• Training iterations (epochs) play a major role in model performance.

• Convolutional networks are a good choice for time series in performance and time.

摘要

•A multiple combination of parameters is better than a single one for deep learning.•Specific validation sets create better models than diverse sets.•Training iterations (epochs) play a major role in model performance.•Convolutional networks are a good choice for time series in performance and time.

论文关键词：Stock returns forecasting,Deep learning,Hyperparameter setting

论文评审过程：Received 25 May 2019, Revised 25 November 2019, Accepted 1 February 2020, Available online 4 February 2020, Version of Record 12 February 2020.

论文官网地址：https://doi.org/10.1016/j.eswa.2020.113271