Introduction

A composite Fear-Greed index combining sentiment, volatility, positioning, and valuation provides holistic assessment of market sentiment. Machine learning weighting—learning optimal weights for combining components—generates more predictive indices than ad-hoc construction. Fear-Greed index predicts reversals: extreme fear precedes bounces, extreme greed precedes crashes.

Composite Index Construction

Components: (1) VIX (volatility); (2) Put/call ratios (hedging demand); (3) Sentiment scores; (4) Valuation multiples (P/E, dividend yield); (5) Momentum. Use ensemble learning to weight components optimally for predicting 1-week equity returns. Backtest weights on historical data.

Trading Application

When Fear-Greed index shows extreme fear (percentile < 10%), buy defensive assets; when extreme greed (percentile > 90%), rotate to cash or shorts. Backtest on 10+ years of data shows profitability in both regimes.

Conclusion

ML-weighted Fear-Greed indices capture multi-modal sentiment signals, improving market sentiment assessment and trading decisions.