Deep Learning for Stock Market Prediction

Area of Science:

• * Computational Finance
• * Data Science
• * Financial Market Analysis

Background:

• * Stock market prediction is complex due to inherent dynamics and non-linearity.
• * Accurate forecasting is crucial for shareholders and investment strategies.
• * Previous models often struggle with the intricate nature of financial markets.

Purpose of the Study:

• * To predict the future values of specific stock market groups.
• * To evaluate the performance of various machine learning algorithms for stock prediction.
• * To assess prediction accuracy for multiple future time horizons.

Main Methods:

• * Utilized historical data for four Tehran Stock Exchange groups: diversified financials, petroleum, non-metallic minerals, and basic metals.
• * Employed machine learning algorithms including Decision Tree, Bagging, Random Forest, Adaptive Boosting (Adaboost), Gradient Boosting, XGBoost, Artificial Neural Networks (ANN), Recurrent Neural Networks (RNN), and Long Short-Term Memory (LSTM).
• * Incorporated ten technical indicators as input features for all prediction models.

Main Results:

• * Long Short-Term Memory (LSTM) achieved the highest accuracy and model fitting ability among all tested algorithms.
• * Tree-based models like Adaboost, Gradient Boosting, and XGBoost showed competitive performance.
• * Predictions were generated for 1, 2, 5, 10, 15, 20, and 30 days in advance.

Conclusions:

• * LSTM is a highly effective model for predicting stock market group values.
• * Advanced ensemble methods and deep learning approaches offer promising results in financial forecasting.
• * The study provides valuable insights into selecting appropriate machine learning techniques for stock market analysis.