Investor Sentiment and Market Movements: A Granger Causality Perspective

Battle Count: 78

The paper establishes a directional causal link (sentiment → price) which is relevant for sentiment-based trading signals. However, the methodology is basic (standard Granger test on daily data), lacks predictive performance metrics, and does not translate findings into actionable trading strategies. Useful as a conceptual validation but limited for direct quantitative implementation.

The pipeline is relatively straightforward: Flair for NLP sentiment scoring (well-documented library) and standard Granger causality test (available in statsmodels/Python). Main complexity lies in data collection (BSE intraday data, curated news headlines) and ensuring stationarity of time series before testing. No custom model architecture or advanced optimization is required.

The methodology is described at a high level (Flair pipeline steps, Granger test steps), but no code, specific hyperparameters, exact dataset source, or software versions are provided. The 80/20 split and lag selection criteria are mentioned but not fully detailed. Reproducing exact results would require access to the same BSE intraday data and news headlines.