Genetic algorithm-optimized multi-channel convolutional neural network for stock market prediction

Hyejung Chung, Kyung shik Shin

Research output: Contribution to journalArticlepeer-review

43 Scopus citations

Abstract

Recently, artificial intelligence technologies have received considerable attention because of their practical applications in various fields. The key factor in this prosperity is deep learning which is inspired by the information processing in biological brains. In this study, we apply one of the representative deep learning techniques multi-channel convolutional neural networks (CNNs) to predict the fluctuation of the stock index. Furthermore, we optimize the network topology of CNN to improve the model performance. CNN has many hyper-parameters that need to be adjusted for constructing an optimal model that can learn the data patterns efficiently. In particular, we focus on the optimization of feature extraction part of CNN, because this is the most important part of the computational procedure of CNN. This study proposes a method to systematically optimize the parameters for the CNN model by using genetic algorithm (GA). To verify the effectiveness of our model, we compare the prediction result with standard artificial neural networks (ANNs) and CNN models. The experimental results show that the GA-CNN outperforms the comparative models and demonstrate the effectiveness of the hybrid approach of GA and CNN.

Original languageEnglish
Pages (from-to)7897-7914
Number of pages18
JournalNeural Computing and Applications
Volume32
Issue number12
DOIs
StatePublished - 1 Jun 2020

Keywords

  • Convolutional neural network
  • Deep learning
  • Genetic algorithm
  • Stock market prediction

Fingerprint

Dive into the research topics of 'Genetic algorithm-optimized multi-channel convolutional neural network for stock market prediction'. Together they form a unique fingerprint.

Cite this