A Seasonal Time-Series Model Based on Gene Expression Programming for Predicting Financial Distress

Area of Science:

• Financial Economics
• Computational Finance
• Data Science

Background:

• Financial distress prediction is a significant challenge in finance.
• Artificial intelligence methods outperform traditional statistical approaches for bankruptcy prediction.
• Company financial data exhibits seasonal, nonlinear, and nonstationary time-series characteristics.

Purpose of the Study:

• To propose a novel nonlinear financial distress prediction model using seasonal time-series gene expression programming.
• To address limitations of previous models by incorporating time-series concepts.
• To develop a model capable of generating interpretable rules and formulas for financial distress.

Main Methods:

• Employed a nonlinear attribute selection method for feature identification.
• Developed a seasonal time-series gene expression programming (STGP) model.
• Integrated attribute selection to reduce data dimensionality and identify core predictive factors.

Main Results:

• The proposed STGP model demonstrated superior performance compared to existing classifiers.
• The model effectively handles seasonal, nonlinear, and nonstationary financial time-series data.
• The integrated attribute selection successfully identified key financial distress indicators.

Conclusions:

• The novel STGP model offers a competitive advantage in financial distress prediction.
• The model provides valuable insights and interpretable rules for investors and decision-makers.
• This approach enhances the accuracy and reliability of financial crisis forecasting.