The 5-Factor IC-Weighted Alpha Model scores each asset across 5 independent alpha factors. Factors are normalized (Z-score) and combined using Information Coefficient (IC) weights — factors that have been more predictive recently get higher weight automatically.
Multi-period ROC: 0.4×ROC_5d + 0.3×ROC_20d + 0.2×ROC_60d + 0.1×ROC_1d
Boosted when Hurst > 0.55 (trending)
−Z-score(Close − SMA_20) / BB_Std
Price far above mean → bearish. Boosted when Hurst < 0.45
0.5×OBV_Slope + 0.3×CMF + 0.2×Volume_Osc
Rising OBV + rising price = institutional accumulation
Monte Carlo Dropout: 20 forward passes → mean prediction + uncertainty
Score = sign(pred − price) × 1/(1 + uncertainty)
Keltner Squeeze detection + ATR percentile rank
Squeeze fires breakout signal aligned with momentum direction