perf: refine the candles window step search

The search space is increased, optuna will takes longer.

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

diff --git a/quickadapter/user_data/config-template.json b/quickadapter/user_data/config-template.json
index e4da3ce86421eacdafea21d29a87be3a02f65661..4c3db3ede3b6dcdd2034c50ffd6db1ab33fddc6a 100644 (file)
--- a/quickadapter/user_data/config-template.json
+++ b/quickadapter/user_data/config-template.json
@@ -119,7 +119,7 @@
         "conv_width": 1,
         "purge_old_models": 2,
         "expiration_hours": 12,
-        "train_period_days": 14,
+        "train_period_days": 30,
         "live_retrain_hours": 0.5,
         "backtest_period_days": 2,
         "write_metrics_to_disk": false,

diff --git a/quickadapter/user_data/freqaimodels/LightGBMRegressorQuickAdapterV35.py b/quickadapter/user_data/freqaimodels/LightGBMRegressorQuickAdapterV35.py
index 91cf2690eea3f0c2aa0923735dbdeb61df2c33d7..bbaccb38fe69988ec2d6312d95c8189b092c161d 100644 (file)
--- a/quickadapter/user_data/freqaimodels/LightGBMRegressorQuickAdapterV35.py
+++ b/quickadapter/user_data/freqaimodels/LightGBMRegressorQuickAdapterV35.py
@@ -223,7 +223,7 @@ def objective(trial, X, y, weights, X_test, y_test, params):
         "reg_lambda": trial.suggest_loguniform("reg_lambda", 1e-8, 10.0),
     }
     params = {**params, **study_params}
-    window = trial.suggest_int("train_period_candles", 1152, 17280, step=300)
+    window = trial.suggest_int("train_period_candles", 1152, 17280, step=100)
 
     # Fit the model
     model = LGBMRegressor(**params)

diff --git a/quickadapter/user_data/freqaimodels/XGBoostRegressorQuickAdapterV35.py b/quickadapter/user_data/freqaimodels/XGBoostRegressorQuickAdapterV35.py
index 082c0bd298eb2b94557f1c9be0513c0f8142871e..6c053d611b6b380b176e6bf0deb9ec3b0c2ac161 100644 (file)
--- a/quickadapter/user_data/freqaimodels/XGBoostRegressorQuickAdapterV35.py
+++ b/quickadapter/user_data/freqaimodels/XGBoostRegressorQuickAdapterV35.py
@@ -231,7 +231,7 @@ def objective(trial, X, y, weights, X_test, y_test, params):
         ],
     }
     params = {**params, **study_params}
-    window = trial.suggest_int("train_period_candles", 1152, 17280, step=300)
+    window = trial.suggest_int("train_period_candles", 1152, 17280, step=100)
 
     # Fit the model
     model = XGBRegressor(**params)