Generic placeholder image

Recent Advances in Computer Science and Communications


ISSN (Print): 2666-2558
ISSN (Online): 2666-2566

Research Article

Time Series Features Extraction and Forecast from Multi-feature Stocks with Hybrid Deep Neural Networks

Author(s): Ye Xu* and Xun Yuan

Volume 14, Issue 8, 2021

Published on: 25 June, 2020

Page: [2662 - 2672] Pages: 11

DOI: 10.2174/2666255813999200625220302

Price: $65


In this paper, we use LSTM and LSTM-CNN models to predict the rise and fall of stock data. It has been proved that LSTM-based models are powerful tools in time series stock data forecast.

Background: Forecasting of time series stock data is important in financial works. Stock data usually have multi-features such as opening price, closing price and so on. The traditional forecast methods, however, are mainly applied to one feature – closing price, or a few, like four or five features. The massive information hidden in the multi-feature data is not thoroughly discovered and used.

Objective: The study aimed to find a method to make use of all information about multi-features and get a forecast model.

Method: LSTM based models are introduced in this paper. For comparison, three models are used, and they are single LSTM model, a hybrid model of LSTM-CNN, and a traditional ARIMA model.

Results: Experiments with different models were performed on stock data with 50 and 230 features, respectively. Results showed that MSE of single LSTM model was 2.4% lower than the ARIMA model and MSE of LSTM-CNN model was 12.57% lower than that of a single LSTM model on 50 features data. On 230 features data, the LSTM-CNN model was found to be improved by 23.41% in forecast accuracy.

Conclusion: In this paper, we used three different models – ARIMA, single LSTM and LSTMCNN hybrid model – to forecast the rise and fall of multi-features stock data. It has been found that the single LSTM model is better than the traditional ARIMA model on average, and the LSTMCNN hybrid model is better than a single LSTM model on 50-feature stock data. Moreover, we used LSTM-CNN model to perform experiments on stock data with 50 and 230 features, respectively and found that the results of the same model on 230 features data were better than that on 50 features data. It has been proved in our work that the LSTM-CNN hybrid model is better than other models and experiments on stock data with more features could result in better outcomes. We will carry out more works on hybrid models next.

Keywords: Multi-feature stock data, LSTM, CNN, time series, forecast, MSE.

Graphical Abstract

Rights & Permissions Print Export Cite as
© 2023 Bentham Science Publishers | Privacy Policy