fix sklearn warnings (#709)

examples/scripts/multi_class_classifier.py

@@ -13,11 +13,6 @@
 from supervised import AutoML
 from supervised.exceptions import AutoMLException
 
-# warnings.filterwarnings('error')
-warnings.filterwarnings(
-    "error", category=pd.core.common.SettingWithCopyWarning
-)  # message="*ndarray*")
-
 df = pd.read_csv("tests/data/iris_missing_values_missing_target.csv")
 X = df[["feature_1", "feature_2", "feature_3", "feature_4"]]
 y = df["class"]

supervised/algorithms/registry.py

@@ -3,7 +3,6 @@
 MULTICLASS_CLASSIFICATION = "multiclass_classification"
 REGRESSION = "regression"
 
-
 class AlgorithmsRegistry:
     registry = {
         BINARY_CLASSIFICATION: {},
@@ -65,3 +64,13 @@ def get_eval_metric(algorithm_name, ml_task, automl_eval_metric):
         return automl_eval_metric
 
 # Import algorithm to be registered
+import supervised.algorithms.baseline
+import supervised.algorithms.catboost
+import supervised.algorithms.decision_tree
+import supervised.algorithms.extra_trees
+import supervised.algorithms.knn
+import supervised.algorithms.lightgbm
+import supervised.algorithms.linear
+import supervised.algorithms.nn
+import supervised.algorithms.random_forest
+import supervised.algorithms.xgboost

supervised/utils/importance.py

@@ -24,7 +24,7 @@ def log_loss_eps(y_true, y_pred):
     return ll
 
 
-log_loss_scorer = make_scorer(log_loss_eps, greater_is_better=False, needs_proba=True)
+log_loss_scorer = make_scorer(log_loss_eps, greater_is_better=False, response_method="predict_proba")
 
 
 class PermutationImportance:

supervised/utils/metric.py

@@ -20,10 +20,8 @@
 
 
 def logloss(y_true, y_predicted, sample_weight=None):
-    epsilon = 1e-6
-    y_predicted = sp.maximum(epsilon, y_predicted)
-    y_predicted = sp.minimum(1 - epsilon, y_predicted)
-    ll = log_loss(y_true, y_predicted, sample_weight=sample_weight)
+    # convert predicted values to float32 to avoid warnings
+    ll = log_loss(y_true, y_predicted.astype(np.float32), sample_weight=sample_weight)
     return ll