Model Extensions
Neural networks are described as inappropriate for this timeframe compared with interpretable linear models. Cross-validation is more statistically rigorous than a single 75/25 split and can provide confidence intervals on signal strength.…
1 sources - 4 claims
Neural networks are described as inappropriate for this timeframe compared with interpretable linear models. Cross-validation is more statistically rigorous than a single 75/25 split and can provide confidence intervals on signal strength. A linear regression can refine the binary mean-reversion signal while remaining interpretable. Regime detection can help stop trading when a crypto signal becomes stale after structural breaks.