CNN-GRUA-FC Stock Price Forecast Model Based on Multi-Factor Analysis

• Published in: Journal of advanced computational intelligence and intelligent informatics Vol. 26; no. 4; pp. 600 - 608
• Main Authors: Yang, Shuying, Guo, Haiming, Li, Junguang
• Format: Journal Article
• Language: English
• Published: Tokyo Fuji Technology Press Co. Ltd 20-07-2022
• Subjects: Accuracy; Algorithms; Artificial neural networks; Datasets; Deep learning; Dynamic characteristics; Factor analysis; Machine learning; Mathematical models; Neural networks; Prediction models; Pricing; Securities markets; Time series; Trends
• ISSN: 1343-0130; 1883-8014
• DOI: 10.20965/jaciii.2022.p0600

To predict stock prices, this paper proposes a CNN-GRUA-FC model based on multi-factor analysis for time series forecasting. First, we use the random forest algorithm to evaluate the importance of the factor series, selecting the factor series of greater importance as the input of the subsequent prediction model. We then use the convolutional neural network (CNN) to extract the spatial characteristics of stock data for prediction, taking advantage of the gated recurrent unit (GRU) neural network to extract the dynamic characteristics of stock data and make prediction. Additionally, we combine an attention mechanism with the GRU neural network (GRUA) to improve its prediction performance. Finally, the prediction results of the two different sub-models are passed through the fully connected (FC) layer to obtain the final predictions. The results show that the prediction accuracy of the CNN-GRUA-FC prediction model proposed in this paper is higher than that of other models.