Added '_BaseEnsemble.recaller' attribute

sklearn2pmml/ensemble/__init__.py

@@ -181,7 +181,7 @@ def predict_proba(self, X, **predict_proba_params):
 
 class _BaseEnsemble(_BaseComposition):
 
-	def __init__(self, steps):
+	def __init__(self, steps, recaller):
 		for step in steps:
 			if type(step) is not tuple:
 				raise TypeError("Step is not a tuple")
@@ -191,6 +191,7 @@ def __init__(self, steps):
 			if not isinstance(predicate, (str, Predicate)):
 				raise TypeError()
 		self.steps = steps
+		self.recaller = recaller
 
 	@property
 	def _steps(self):
@@ -207,6 +208,11 @@ def set_params(self, **kwargs):
 		self._set_params("_steps", **kwargs)
 		return self
 
+	def _to_evaluation_dataset(self, X):
+		if self.recaller is not None:
+			return self.recaller.transform(X)
+		return X
+
 def _to_sparse(X, step_mask, step_result):
 	# Make array
 	if len(step_result.shape) == 1:
@@ -256,18 +262,19 @@ def augment(self, X):
 
 class EstimatorChain(_BaseEnsemble):
 
-	def __init__(self, steps, multioutput = True):
-		super(EstimatorChain, self).__init__(steps)
+	def __init__(self, steps, recaller = None, multioutput = True):
+		super(EstimatorChain, self).__init__(steps, recaller)
 		self.multioutput = multioutput
 
 	def fit(self, X, y, **fit_params):
 		if len(y.shape) > 1:
 			if len(self.steps) != y.shape[1]:
 				raise ValueError()
 			y = numpy.asarray(y)
+		X_eval = self._to_evaluation_dataset(X)
 		i = 0
 		for name, estimator, predicate in self.steps:
-			step_mask = eval_expr_rows(X, predicate, dtype = bool)
+			step_mask = eval_expr_rows(X_eval, predicate, dtype = bool)
 			if numpy.sum(step_mask) < 1:
 				raise ValueError(predicate)
 			step_X = X[step_mask]
@@ -286,8 +293,9 @@ def fit(self, X, y, **fit_params):
 
 	def _predict(self, X, predict_method):
 		result = None
+		X_eval = self._to_evaluation_dataset(X)
 		for name, estimator, predicate in self.steps:
-			step_mask = eval_expr_rows(X, predicate, dtype = bool)
+			step_mask = eval_expr_rows(X_eval, predicate, dtype = bool)
 			if numpy.sum(step_mask) < 1:
 				continue
 			step_X = X[step_mask]
@@ -316,13 +324,14 @@ def predict_proba(self, X):
 
 class SelectFirstEstimator(_BaseEnsemble):
 
-	def __init__(self, steps):
-		super(SelectFirstEstimator, self).__init__(steps)
+	def __init__(self, steps, recaller):
+		super(SelectFirstEstimator, self).__init__(steps, recaller)
 
 	def fit(self, X, y, **fit_params):
+		X_eval = self._to_evaluation_dataset(X)
 		mask = numpy.zeros(X.shape[0], dtype = bool)
 		for name, estimator, predicate in self.steps:
-			step_mask = eval_expr_rows(X, predicate, dtype = bool)
+			step_mask = eval_expr_rows(X_eval, predicate, dtype = bool)
 			step_mask[mask] = False
 			if numpy.sum(step_mask) < 1:
 				raise ValueError(predicate)
@@ -335,9 +344,10 @@ def fit(self, X, y, **fit_params):
 
 	def _predict(self, X, predict_method):
 		result = None
+		X_eval = self._to_evaluation_dataset(X)
 		mask = numpy.zeros(X.shape[0], dtype = bool)
 		for name, estimator, predicate in self.steps:
-			step_mask = eval_expr_rows(X, predicate, dtype = bool)
+			step_mask = eval_expr_rows(X_eval, predicate, dtype = bool)
 			step_mask[mask] = False
 			if numpy.sum(step_mask) < 1:
 				continue
@@ -359,13 +369,13 @@ def predict(self, X):
 
 class SelectFirstRegressor(SelectFirstEstimator, RegressorMixin):
 
-	def __init__(self, steps):
-		super(SelectFirstRegressor, self).__init__(steps)
+	def __init__(self, steps, recaller = None):
+		super(SelectFirstRegressor, self).__init__(steps, recaller)
 
 class SelectFirstClassifier(SelectFirstEstimator, ClassifierMixin):
 
-	def __init__(self, steps):
-		super(SelectFirstClassifier, self).__init__(steps)
+	def __init__(self, steps, recaller = None):
+		super(SelectFirstClassifier, self).__init__(steps, recaller)
 
 	def predict_proba(self, X):
 		return self._predict(X, "predict_proba")