diff --git a/rosie/chamber_of_deputies/classifiers/meal_price_outlier_classifier.py b/rosie/chamber_of_deputies/classifiers/meal_price_outlier_classifier.py
index 8620271..9e9cf55 100644
--- a/rosie/chamber_of_deputies/classifiers/meal_price_outlier_classifier.py
+++ b/rosie/chamber_of_deputies/classifiers/meal_price_outlier_classifier.py
@@ -28,6 +28,11 @@ class MealPriceOutlierClassifier(TransformerMixin):
 
     HOTEL_REGEX = r'hote(?:(?:ls?)|is)'
     CLUSTER_KEYS = ['mean', 'std']
+    COLS = ['applicant_id',
+            'category',
+            'net_value',
+            'recipient',
+            'recipient_id']
 
     def fit(self, X):
         _X = X[self.__applicable_rows(X)]
@@ -49,7 +54,7 @@ def transform(self, X=None):
         pass
 
     def predict(self, X):
-        _X = X.copy()
+        _X = X[self.COLS].copy()
         companies = _X[self.__applicable_rows(_X)] \
             .groupby('recipient_id').apply(self.__company_stats) \
             .reset_index()
@@ -68,7 +73,7 @@ def predict(self, X):
         _X = pd.merge(_X, known_thresholds, how='left')
         if 'cnpj_threshold' in _X.columns:
             _X.loc[_X['cnpj_threshold'].notnull(),
-                  'threshold'] = _X['cnpj_threshold']
+                   'threshold'] = _X['cnpj_threshold']
         _X['y'] = 1
         is_outlier = self.__applicable_rows(_X) & \
             _X['threshold'].notnull() & \
diff --git a/rosie/chamber_of_deputies/classifiers/monthly_subquota_limit_classifier.py b/rosie/chamber_of_deputies/classifiers/monthly_subquota_limit_classifier.py
index af60a39..0c74925 100644
--- a/rosie/chamber_of_deputies/classifiers/monthly_subquota_limit_classifier.py
+++ b/rosie/chamber_of_deputies/classifiers/monthly_subquota_limit_classifier.py
@@ -12,15 +12,30 @@ class MonthlySubquotaLimitClassifier(TransformerMixin):
     issue_date : datetime column
         Date when the expense was made.
 
+    month : int column
+        The quota month matching the expense request.
+
     net_value : float column
         The value of the expense.
+
+    subquota_number : category column
+        A number to classify a category of expenses.
+
+    year : int column
+        The quota year matching the expense request.
     """
 
     KEYS = ['applicant_id', 'month', 'year']
+    COLS = ['applicant_id',
+            'issue_date',
+            'month',
+            'net_value',
+            'subquota_number',
+            'year']
 
     def fit(self, X):
         self.X = X
-        self._X = self.X.copy()
+        self._X = self.X[self.COLS].copy()
         self.__create_columns()
         return self
 
diff --git a/rosie/chamber_of_deputies/classifiers/traveled_speeds_classifier.py b/rosie/chamber_of_deputies/classifiers/traveled_speeds_classifier.py
index dd7529f..6bd5c86 100644
--- a/rosie/chamber_of_deputies/classifiers/traveled_speeds_classifier.py
+++ b/rosie/chamber_of_deputies/classifiers/traveled_speeds_classifier.py
@@ -36,6 +36,12 @@ class TraveledSpeedsClassifier(TransformerMixin):
     """
 
     AGG_KEYS = ['applicant_id', 'issue_date']
+    COLS = ['applicant_id',
+            'category',
+            'is_party_expense',
+            'issue_date',
+            'latitude',
+            'longitude']
 
     def __init__(self, contamination=.001):
         if contamination in [0, 1]:
@@ -57,7 +63,7 @@ def transform(self, X=None):
     def predict(self, X):
         check_is_fitted(self, ['polynomial', '_polynomial_fn'])
 
-        _X = X.copy()
+        _X = X[self.COLS].copy()
         _X = self.__aggregate_dataset(_X)
         _X = self.__classify_dataset(_X)
         _X = pd.merge(X, _X, how='left', left_on=self.AGG_KEYS, right_on=self.AGG_KEYS)