Harnessing ChatGPT for predictive financial factor generation: A new frontier in financial analysis and forecasting

Abstract

The search for predictive financial factors in stock pricing of companies has long been a key focus in accounting and finance, but traditional methods often require complex, subjective inputs. This paper introduces a method using ChatGPT-4 to generate financial factors based on the structure of financial statements and key variables, eliminating the need for numerical data. Leveraging GPT’s natural language processing capabilities and extensive knowledge base, our approach efficiently generates factors that are highly predictive of future returns and exhibit robustness over time, unaffected by variations in different conversational windows. Regression analysis demonstrates that these factors cannot be linearly explained by traditional financial factors. This paper highlights AI’s potential in revolutionizing financial analysis and decision-making.