Enhancing the pricing efficiency of financial assets with an optimized bayesian network based on efficient fusion

Area of Science:

• Quantitative Finance
• Machine Learning
• Financial Econometrics

Background:

• Traditional financial asset pricing models struggle with accuracy and adaptability in high-frequency trading environments.
• Limitations include difficulties in capturing complex sequential dependencies and hierarchical market structures.

Purpose of the Study:

• To develop an advanced financial asset pricing model for high-frequency trading.
• To enhance accuracy, adaptability, and interpretability by integrating deep learning and probabilistic graphical models.

Main Methods:

• A Transformer-based Efficiently-Fused Optimized Bayesian Network (Trans-EFOBN) was developed.
• The framework combines a masked transformer for sequential feature extraction with temporal logic constraints.
• A Dynamic Bayesian Network (DBN) establishes hierarchical dependencies between macro and micro market variables.

Main Results:

• The Trans-EFOBN model achieved a Mean Absolute Error (MAE) of 0.037, a 25% reduction compared to baseline models.
• The model attained an R² of 0.86, indicating strong explanatory power.
• Simulated trading yielded a 14.2% annualized return and a Sharpe ratio of 0.95.

Conclusions:

• Integrating structural dependency logic with dynamic probabilistic inference enhances asset pricing efficiency.
• The Trans-EFOBN model offers robust technical support for high-frequency quantitative trading.
• The approach demonstrates superior performance and interpretability over traditional methods.