Symbolic Modeling for financial asset pricing
This paper presents a unified Symbolic Modeling approach leveraging side information and genetic programming to improve financial asset pricing models.
What it examines
The paper introduces Symbolic Modeling, an extension of Symbolic Regression that generates a unified asset pricing model for multiple datasets. It uses side information, genetic programming, and reinforcement learning to develop general functions with adaptable coefficients, aiming to improve predictive accuracy over classic models.
What it concludes
The study shows that the unified Symbolic Modeling approach reduces prediction error and abnormal returns compared to CAPM and FF-3. Applications include enhanced financial analysis, better investment decisions, and potential use in other fields requiring clear, adaptable mathematical models.
Evidence objects
Researchers introduce Symbolic Modeling, a breakthrough generalization of Symbolic Regression that unifies robust asset pricing models across diverse datasets, effectively reducing complexity and overfitting while seamlessly integrating valuable side information.
key_findings bullet 1 · key_findings · validation V0
The SIBSR framework combines side information with genetic programming and reinforcement learning, attaining lower normalized $\mathrm{RMSE}$ and absolute $\alpha$ values than traditional CAPM and Fama-French 3-Factor models across rigorous testing.
key_findings bullet 2 · key_findings · validation V0
Authors propose a formulation that replaces rigid symbolic regression with a unified model allowing dataset-specific tuning; tests on 365 companies over four decades prove interpretability, though computational demands hinder adoption.
key_findings bullet 3 · key_findings · validation V0
Integrating symbolic regression with side information boosting and reinforcement learning for financial asset pricing, the paper introduces a novel $SIBSR$ process and model-level GP bundling. It creatively extends established genetic programming techniques, offering a fresh, compelling perspective that promises significant impact, particularly for quantitative finance research and computational asset valuation.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
- Symbolic Regression is a machine learning technique that discovers an unknown function from its samples. Compared to conventional regression techniques (eg, linear …
Source row: 1888 · abstract type: snippet