perf(qav3): fine tune labeling optimization

Signed-off-by: Jérôme Benoit <jerome.benoit@piment-noir.org>

diff --git a/quickadapter/user_data/freqaimodels/QuickAdapterRegressorV3.py b/quickadapter/user_data/freqaimodels/QuickAdapterRegressorV3.py
index 99d52958a86335f75b8efed96a8398ebe971edd0..2d00b02b66df6fb9305c9cc24f1845c5b618cb6a 100644 (file)
--- a/quickadapter/user_data/freqaimodels/QuickAdapterRegressorV3.py
+++ b/quickadapter/user_data/freqaimodels/QuickAdapterRegressorV3.py
@@ -1420,7 +1420,7 @@ def label_objective(
         max_label_period_candles,
         step=candles_step,
     )
-    label_natr_ratio = trial.suggest_float("label_natr_ratio", 6.0, 36.0, step=0.01)
+    label_natr_ratio = trial.suggest_float("label_natr_ratio", 7.0, 35.0, step=0.01)
 
     label_period_cycles = fit_live_predictions_candles / label_period_candles
     df = df.iloc[-(max(2, int(label_period_cycles)) * label_period_candles) :]