Introduction

Ensemble sentiment—combining multiple independent sentiment sources (news, social media, options, holdings)—often outperforms individual sources. The wisdom-of-crowds effect: aggregated opinions are smarter than individuals. Machine learning quantifies this by comparing ensemble sentiment predictions to individual source predictions for stock returns.

Ensemble Sentiment Construction

Combine: (1) News sentiment; (2) Social media sentiment; (3) Options implied sentiment (put/call ratios); (4) Insider trading sentiment; (5) Institutional flow sentiment. Weight sources by past predictive power. Aggregate into ensemble score. Compare prediction accuracy to individual sources.

Results

Ensemble sentiment predicts stock returns more accurately than any individual source. Ensemble correlates with 1-month forward returns at 0.12 (modest but significant). Ensemble-based trading strategy beats single-source strategies in risk-adjusted returns.

Conclusion

Ensemble sentiment captures wisdom-of-crowds effect, enabling more reliable sentiment-based trading.