Stock Market Forecasting Using Restricted Gene Expression Programming.

Area of Science:

• Computational Finance
• Artificial Intelligence
• Time Series Analysis

Background:

• Stock index prediction is a challenging task with significant financial implications.
• Existing methods often struggle with accuracy and convergence speed.
• The S-system model offers a promising framework for complex system modeling.

Purpose of the Study:

• To propose a novel stock index prediction method using an S-system model.
• To enhance the S-system model's structure and parameter optimization.
• To validate the proposed method's performance against established prediction techniques.

Main Methods:

• Restricted Gene Expression Programming (RGEP) for S-system structure optimization.
• A hybrid Brain Storm Optimization (BSO) and Particle Swarm Optimization (PSO) algorithm for S-system parameter tuning.
• Validation using five major stock market indices: Dow Jones, Hang Seng, NASDAQ, Shanghai Composite, and SZSE Component.

Main Results:

• The proposed S-system model achieved superior performance in 1-week-ahead and 1-month-ahead stock index prediction.
• Outperformed Deep Recurrent Neural Network (DRNN), Flexible Neural Tree (FNT), Radial Basis Function (RBF), Backpropagation (BP), and ARIMA models.
• The hybrid BSO-PSO algorithm exhibited faster convergence compared to standalone PSO and BSO.

Conclusions:

• The novel S-system based approach provides an effective solution for accurate stock index prediction.
• The hybrid optimization algorithm significantly improves model parameter tuning efficiency.
• This research contributes a robust and efficient method for financial market forecasting.