Abstract:Numerous studies are based on the correlation among stock factors, which affects the measurement value and interpretability of such studies. Research on the causality among stock factors primarily relies on statistical models and machine learning algorithms, thereby failing to fully exploit the formidable computational capabilities of deep learning models. Moreover, the inference of causal relationships largely depends on the Granger causality test, which is not suitable for non-stationary and non-linear stock factors. Also, most existing studies do not consider the impact of confounding variables or further validation of causal relationships. In response to the current research deficiencies, this paper introduces a deep learning-based algorithm aimed at inferring causal relationships between stock closing prices and relevant factors. To achieve this, causal diagrams from the structural causal model (SCM) were integrated into the analysis of stock data. Subsequently, a sliding window strategy combined with Gated Recurrent Units (GRUs) was employed to predict the potential values of closing prices, and a grouped architecture was constructed inspired by the Potential Outcomes Framework (POF) for controlling confounding variables. The architecture was employed to infer causal relationships between closing price and relevant factors through the non-linear Granger causality test. Finally, comparative experimental results demonstrate a marked enhancement in the accuracy and performance of closing price predictions when causal factors were incorporated into the prediction model. This finding not only validates the correctness of the causal inference, but also strengthens the reliability and validity of the proposed methodology. Consequently, this study has significant practical implications for the analysis of causality in financial time series data and the prediction of stock prices.

Causal Factor Investing: Can Factor Investing Become Scientific?

Scientific Approach : Causal Factor Investing : Causal Adjusted Randomized Factor Investing

Factor Investing for the Long Run

Navigating the factor zoo around the world: an institutional investor perspective

The Evolving Causal Structure of Equity Risk Factors

Exploring the merits of five-factor investing: A critical literature review

Factor Investing with a Deep Multi-Factor Model

Improving Estimation of Portfolio Risk Using New Statistical Factors

Factor Investing in Credit

Anomaly or Risk Factor? A Stepwise Evaluation

Automatic Linear Causal Relationship Identification for Financial Factor Modeling

Causal Analysis of Generic Time Series Data Applied for Market Prediction

Deep Learning-Based Causal Inference Architecture and Algorithm between Stock Closing Price and Relevant Factors

Trading with Time Series Causal Discovery: An Empirical Study

Factor Models for Portfolio Selection in Large Dimensions: The Good, the Better and the Ugly

Lucky factors

Sector-Based Factor Models for Asset Returns

Factor investing: alpha concentration versus diversification

Epistemic Limits of Empirical Finance: Causal Reductionism and Self-Reference

Non-parametric causality detection: An application to social media and financial data

Causal Inference in Finance: An Expertise-Driven Model for Instrument Variables Identification and Interpretation