Forecasting stock prices with a feature fusion LSTM-CNN model using different representations of the same data

Taewook Kim, Ha Young Kim

Research output: Contribution to journalArticle

8 Citations (Scopus)

Abstract

Forecasting stock prices plays an important role in setting a trading strategy or determining the appropriate timing for buying or selling a stock. We propose a model, called the feature fusion long short-term memory-convolutional neural network (LSTM-CNN) model, that combines features learned from different representations of the same data, namely, stock time series and stock chart images, to predict stock prices. The proposed model is composed of LSTM and a CNN, which are utilized for extracting temporal features and image features. We measure the performance of the proposed model relative to those of single models (CNN and LSTM) using SPDR S&P 500 ETF data. Our feature fusion LSTM-CNN model outperforms the single models in predicting stock prices. In addition, we discover that a candlestick chart is the most appropriate stock chart image to use to forecast stock prices. Thus, this study shows that prediction error can be efficiently reduced by using a combination of temporal and image features from the same data rather than using these features separately.

Original languageEnglish
Article numbere0212320
JournalPloS one
Volume14
Issue number2
DOIs
Publication statusPublished - 2019 Feb

Fingerprint

Neural Networks (Computer)
Long-Term Memory
Short-Term Memory
neural networks
Fusion reactions
Neural networks
purchasing
Long short-term memory
Time series
time series analysis
Sales
prediction

All Science Journal Classification (ASJC) codes

  • Biochemistry, Genetics and Molecular Biology(all)
  • Agricultural and Biological Sciences(all)
  • General

Cite this

@article{98c3de116c2d4966b7db0cec0736beed,
title = "Forecasting stock prices with a feature fusion LSTM-CNN model using different representations of the same data",
abstract = "Forecasting stock prices plays an important role in setting a trading strategy or determining the appropriate timing for buying or selling a stock. We propose a model, called the feature fusion long short-term memory-convolutional neural network (LSTM-CNN) model, that combines features learned from different representations of the same data, namely, stock time series and stock chart images, to predict stock prices. The proposed model is composed of LSTM and a CNN, which are utilized for extracting temporal features and image features. We measure the performance of the proposed model relative to those of single models (CNN and LSTM) using SPDR S&P 500 ETF data. Our feature fusion LSTM-CNN model outperforms the single models in predicting stock prices. In addition, we discover that a candlestick chart is the most appropriate stock chart image to use to forecast stock prices. Thus, this study shows that prediction error can be efficiently reduced by using a combination of temporal and image features from the same data rather than using these features separately.",
author = "Taewook Kim and Kim, {Ha Young}",
year = "2019",
month = "2",
doi = "10.1371/journal.pone.0212320",
language = "English",
volume = "14",
journal = "PLoS One",
issn = "1932-6203",
publisher = "Public Library of Science",
number = "2",

}

Forecasting stock prices with a feature fusion LSTM-CNN model using different representations of the same data. / Kim, Taewook; Kim, Ha Young.

In: PloS one, Vol. 14, No. 2, e0212320, 02.2019.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Forecasting stock prices with a feature fusion LSTM-CNN model using different representations of the same data

AU - Kim, Taewook

AU - Kim, Ha Young

PY - 2019/2

Y1 - 2019/2

N2 - Forecasting stock prices plays an important role in setting a trading strategy or determining the appropriate timing for buying or selling a stock. We propose a model, called the feature fusion long short-term memory-convolutional neural network (LSTM-CNN) model, that combines features learned from different representations of the same data, namely, stock time series and stock chart images, to predict stock prices. The proposed model is composed of LSTM and a CNN, which are utilized for extracting temporal features and image features. We measure the performance of the proposed model relative to those of single models (CNN and LSTM) using SPDR S&P 500 ETF data. Our feature fusion LSTM-CNN model outperforms the single models in predicting stock prices. In addition, we discover that a candlestick chart is the most appropriate stock chart image to use to forecast stock prices. Thus, this study shows that prediction error can be efficiently reduced by using a combination of temporal and image features from the same data rather than using these features separately.

AB - Forecasting stock prices plays an important role in setting a trading strategy or determining the appropriate timing for buying or selling a stock. We propose a model, called the feature fusion long short-term memory-convolutional neural network (LSTM-CNN) model, that combines features learned from different representations of the same data, namely, stock time series and stock chart images, to predict stock prices. The proposed model is composed of LSTM and a CNN, which are utilized for extracting temporal features and image features. We measure the performance of the proposed model relative to those of single models (CNN and LSTM) using SPDR S&P 500 ETF data. Our feature fusion LSTM-CNN model outperforms the single models in predicting stock prices. In addition, we discover that a candlestick chart is the most appropriate stock chart image to use to forecast stock prices. Thus, this study shows that prediction error can be efficiently reduced by using a combination of temporal and image features from the same data rather than using these features separately.

UR - http://www.scopus.com/inward/record.url?scp=85061557918&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85061557918&partnerID=8YFLogxK

U2 - 10.1371/journal.pone.0212320

DO - 10.1371/journal.pone.0212320

M3 - Article

C2 - 30768647

AN - SCOPUS:85061557918

VL - 14

JO - PLoS One

JF - PLoS One

SN - 1932-6203

IS - 2

M1 - e0212320

ER -