Added Additional Regression Objectives

docs/source/api_reference.rst

@@ -130,6 +130,12 @@ Regression
     :nosignatures:
 
     R2
+    MAE
+    MSE
+    MSLE
+    MedianAE
+    MaxError
+    ExpVariance
 
 .. currentmodule:: evalml.problem_types
 
@@ -155,6 +161,3 @@ Tuners
     :nosignatures:
 
     SKOptTuner
-
-
-

evalml/objectives/__init__.py

@@ -2,8 +2,10 @@
 from .fraud_cost import FraudCost
 from .lead_scoring import LeadScoring
 from .standard_metrics import (
-    F1, F1Micro, F1Macro, F1Weighted, Precision, PrecisionMicro, PrecisionMacro, PrecisionWeighted, Recall, RecallMicro, RecallMacro, RecallWeighted,
-    AUC, AUCMicro, AUCMacro, AUCWeighted, LogLoss, MCC, R2
+    AUC, AUCMacro, AUCMicro, AUCWeighted, ExpVariance, F1, F1Macro, F1Micro,
+    F1Weighted, LogLoss, MAE, MaxError, MCC, MedianAE, MSE, MSLE, Precision,
+    PrecisionMacro, PrecisionMicro, PrecisionWeighted, R2, Recall, RecallMacro,
+    RecallMicro, RecallWeighted
     )
 from .objective_base import ObjectiveBase
 from .utils import get_objective, get_objectives

evalml/objectives/standard_metrics.py

@@ -239,6 +239,78 @@ def score(self, y_predicted, y_true):
         return metrics.r2_score(y_true, y_predicted)
 
 
+class MAE(ObjectiveBase):
+    """Mean absolute error for regression"""
+    needs_fitting = False
+    greater_is_better = False
+    need_proba = False
+    name = "MAE"
+    problem_types = [ProblemTypes.REGRESSION]
+
+    def score(self, y_predicted, y_true):
+        return metrics.mean_absolute_error(y_true, y_predicted)
+
+
+class MSE(ObjectiveBase):
+    """Mean squared error for regression"""
+    needs_fitting = False
+    greater_is_better = False
+    need_proba = False
+    name = "MSE"
+    problem_types = [ProblemTypes.REGRESSION]
+
+    def score(self, y_predicted, y_true):
+        return metrics.mean_squared_error(y_true, y_predicted)
+
+
+class MSLE(ObjectiveBase):
+    """Mean squared log error for regression"""
+    needs_fitting = False
+    greater_is_better = False
+    need_proba = False
+    name = "MSLE"
+    problem_types = [ProblemTypes.REGRESSION]
+
+    def score(self, y_predicted, y_true):
+        return metrics.mean_squared_log_error(y_true, y_predicted)
+
+
+class MedianAE(ObjectiveBase):
+    """Median absolute error for regression"""
+    needs_fitting = False
+    greater_is_better = False
+    need_proba = False
+    name = "MedAE"
+    problem_types = [ProblemTypes.REGRESSION]
+
+    def score(self, y_predicted, y_true):
+        return metrics.median_absolute_error(y_true, y_predicted)
+
+
+class MaxError(ObjectiveBase):
+    """Maximum residual error for regression"""
+    needs_fitting = False
+    greater_is_better = False
+    need_proba = False
+    name = "MaxError"
+    problem_types = [ProblemTypes.REGRESSION]
+
+    def score(self, y_predicted, y_true):
+        return metrics.max_error(y_true, y_predicted)
+
+
+class ExpVariance(ObjectiveBase):
+    """Explained variance score for regression"""
+    needs_fitting = False
+    greater_is_better = True
+    need_proba = False
+    name = "ExpVariance"
+    problem_types = [ProblemTypes.REGRESSION]
+
+    def score(self, y_predicted, y_true):
+        return metrics.explained_variance_score(y_true, y_predicted)
+
+
 def _handle_predictions(y_true, y_pred):
     if len(np.unique(y_true)) > 2:
         classes = np.unique(y_true)

evalml/objectives/utils.py

@@ -23,6 +23,12 @@
     "log_loss": standard_metrics.LogLoss(),
     "mcc": standard_metrics.MCC(),
     "r2": standard_metrics.R2(),
+    "mae": standard_metrics.MAE(),
+    "mse": standard_metrics.MSE(),
+    "msle": standard_metrics.MSLE(),
+    "median_ae": standard_metrics.MedianAE(),
+    "max_error": standard_metrics.MaxError(),
+    "exp_var": standard_metrics.ExpVariance()
 }
 
 

evalml/tests/objective_tests/test_objectives.py

@@ -18,7 +18,7 @@ def test_get_objective():
 def test_get_objectives_types():
     assert len(get_objectives('multiclass')) == 14
     assert len(get_objectives('binary')) == 6
-    assert len(get_objectives('regression')) == 1
+    assert len(get_objectives('regression')) == 7
 
 
 def test_binary_average(X_y):