AI Stock Analysis

This study explores how the tone of financial news headlines aligns with movements in the Dow Jones Industrial Average (DIA). Headlines are aggregated per day, cleaned, and scored with the VADER sentiment analyzer so we can observe whether collective optimism or pessimism shows up in the market tape. In parallel, we pull historical DIA prices via yfinance, compute daily returns, and synchronize both datasets to form a single timeline of sentiment versus performance.

A Python notebook powers the workflow: pandas handles preprocessing, VADER delivers positive/negative/ neutral scores, and matplotlib highlights the relationships through scatterplots and distribution charts. We also support lag analysis—yesterday’s sentiment plotted against today’s return—to inspect whether news leads price action by a session or two. The combined feature set turns messy headlines into structured, analysis-ready signals without leaving the scripting environment.

Initial findings show sentiment values clustering around neutral, reminding us to watch for noisy inputs and outliers produced by empty headlines. Correlations remain modest, yet they tighten when we inspect lagged signals, suggesting room for predictive strategies or downstream trading rules. Future work will scale the approach to additional indices, experiment with finance-tuned language models like FinBERT, and expand into weekly or monthly horizons to capture slower-moving macro narratives.