Research on Classification-Based Financial Market Prediction Algorithms Using Deep Learning
Download as PDF
DOI: 10.23977/GEFHR2020.033
Author(s)
Wang Yimeng, Chen Miao, Yu Junqi
Corresponding Author
Wang Yimeng
ABSTRACT
Deep neural networks (DNNs) are a powerful artificial neural network (ANN) using multiple hidden layers. In recent years, it has gained considerable attention in the fields of speech conversion and image recognition, because of their superior the predictive properties of the algorithm include robustness to overfitting. However, their application in algorithmic trading has not been studied before, partly because of their computational complexity. Application. Specifically, we describe the configuration and training methods, and then demonstrate their application of a reverse test of a simple trading strategy at 43-minute intervals on 43 different commodities and foreign exchange futures. All results are generated using the C++ implementation on the Intel Xeon Phi coprocessor (11.4 times faster than the serial version) and the Python policy back testing environment, both of which are open source code written by the author.
KEYWORDS
Financial Market, Prediction Algorithms, Deep Learning