A cooperative deep learning model for stock market prediction using deep autoencoder and sentiment analysis

Area of Science:

• * Computational Finance
• * Artificial Intelligence in Finance
• * Machine Learning for Financial Markets

Background:

• * Stock market prediction is complex due to market volatility.
• * Traditional machine learning models struggle with accuracy.
• * Integrating news sentiment analysis can improve prediction performance.

Purpose of the Study:

• * To develop a cooperative deep-learning architecture for stock market prediction.
• * To effectively denoise stock data and incorporate news sentiment.
• * To achieve higher prediction accuracy compared to existing models.

Main Methods:

• * A deep autoencoder was used for denoising historical stock data.
• * Lexicon-based software performed sentiment analysis on news headlines.
• * Long Short-Term Memory (LSTM)/Gated Recurrent Unit (GRU) layers were employed for prediction, integrating denoised data and sentiment scores.

Main Results:

• * The proposed model, combining a deep autoencoder with news sentiments, outperformed standalone LSTM/GRU models.
• * The cooperative architecture demonstrated superior performance in stock market prediction.
• * Results favorably compared with current state-of-the-art models in the field.

Conclusions:

• * The cooperative deep-learning architecture effectively addresses noise in stock data and captures market sentiment.
• * Integrating news sentiment significantly enhances stock market prediction accuracy.
• * The developed model offers a promising advancement for financial market forecasting.