A thinking innovation strategy based Northern goshawk optimizer enhanced extreme learning machine for bankruptcy...

Area of Science:

• Financial Risk Management
• Computational Intelligence
• Machine Learning

Background:

• Bankruptcy risk prediction is crucial for financial institutions.
• Traditional models face challenges with high-dimensional financial data.
• Metaheuristic algorithms offer potential for enhanced model performance.

Purpose of the Study:

• To propose a novel bankruptcy prediction model using an improved metaheuristic algorithm.
• To enhance the performance of Kernel Extreme Learning Machine (KELM) for financial risk assessment.
• To introduce the Thought-Inspired Strategy Northern goshawk Optimizer (TIS_NGO) for parameter optimization.

Main Methods:

• Integration of KELM with the TIS_NGO algorithm.
• TIS_NGO enhancements include divergence-based innovation, differential evolution-inspired prey-attacking, and opposition-based boundary control.
• Model performance evaluated on CEC2017, CEC2022 benchmark suites, and the Wieslaw bankruptcy dataset.

Main Results:

• TIS_NGO demonstrated superior convergence speed and solution accuracy compared to standard NGO, PSO, and GWO.
• The TIS_NGO-optimized KELM achieved high classification accuracy and robustness on the bankruptcy dataset.
• Validated the effectiveness of combining advanced metaheuristics with machine learning for financial forecasting.

Conclusions:

• The proposed TIS_NGO-KELM model offers a promising advancement in bankruptcy prediction accuracy and stability.
• This approach provides a new technical pathway for early warning systems in the financial domain.
• Highlights the potential of improved metaheuristic algorithms in financial risk management.