Leveraging artificial intelligence models for financial forecasting

This study investigates the predictive validity of generative artificial intelligence (AI) in financial forecasting. Specifically, it evaluates the zero-shot forecasting capabilities of GPT-4o and Claude Sonnet 3.5 by comparing their predicted stock prices against actual closing prices from a cross-industry portfolio as of February 3, 2025. Utilizing standardized statistical measures such as mean absolute error (MAE), root mean squared error (RMSE), mean absolute percentage error (MAPE), correlation coefficients, and R², the study finds that Claude Sonnet 3.5 consistently outperforms GPT-4o in predictive accuracy and correlation. The research also examines directional bias and sector-specific performance.