From 53af39d42af8ac3063d5a804101e6bffdaf1dc3f Mon Sep 17 00:00:00 2001
From: GUI <g.navas.palencia@gmail.com>
Date: Wed, 13 Apr 2022 16:20:41 +0200
Subject: [PATCH] integer intercept if rounding=True

---
 optbinning/scorecard/scorecard.py | 30 ++++++++++++++++++++++++++----
 tests/test_scorecard.py           | 22 +++++++++++++++++++++-
 2 files changed, 47 insertions(+), 5 deletions(-)

diff --git a/optbinning/scorecard/scorecard.py b/optbinning/scorecard/scorecard.py
index 2afd49d..6e36091 100644
--- a/optbinning/scorecard/scorecard.py
+++ b/optbinning/scorecard/scorecard.py
@@ -61,22 +61,26 @@ def _check_parameters(binning_process, estimator, scaling_method,
     if not isinstance(rounding, bool):
         raise TypeError("rounding must be a boolean; got {}.".format(rounding))
 
+    if rounding and scaling_method is None:
+        raise ValueError("rounding is only applied if scaling method is "
+                         "not None.")
+
     if not isinstance(verbose, bool):
         raise TypeError("verbose must be a boolean; got {}.".format(verbose))
 
 
 def _check_scorecard_scaling(scaling_method, scaling_method_params,
-                             target_type):
+                             rounding, target_type):
     if scaling_method is not None:
         if scaling_method == "pdo_odds":
-            default_keys = ["pdo", "odds", "scorecard_points"]
+            default_keys = ("pdo", "odds", "scorecard_points")
 
             if target_type != "binary":
                 raise ValueError('scaling_method "pd_odds" is not supported '
                                  'for a continuous target.')
 
         elif scaling_method == "min_max":
-            default_keys = ["min", "max"]
+            default_keys = ("min", "max")
 
         if set(scaling_method_params.keys()) != set(default_keys):
             raise ValueError("scaling_method_params must be {} given "
@@ -102,6 +106,17 @@ def _check_scorecard_scaling(scaling_method, scaling_method_params,
                                  .format(scaling_method_params["min"],
                                          scaling_method_params["max"]))
 
+            if rounding:
+                score_min = scaling_method_params['min']
+                if int(score_min) != score_min:
+                    raise ValueError("min must be an integer if rounding=True"
+                                     "; got {}.".format(score_min))
+
+                score_max = scaling_method_params['max']
+                if int(score_max) != score_max:
+                    raise ValueError("max must be an integer if rounding=True"
+                                     "; got {}.".format(score_max))
+
 
 def _compute_scorecard_points(points, binning_tables, method, method_data,
                               intercept, reverse_scorecard):
@@ -483,6 +498,7 @@ def _fit(self, X, y, sample_weight, metric_special, metric_missing,
 
         _check_scorecard_scaling(self.scaling_method,
                                  self.scaling_method_params,
+                                 self.rounding,
                                  self._target_dtype)
 
         # Check sample weight
@@ -595,8 +611,11 @@ def _fit(self, X, y, sample_weight, metric_special, metric_missing,
         time_rounding = time.perf_counter()
         if self.rounding:
             points = df_scorecard["Points"]
-            if self.scaling_method == "pdo_odds":
+            if self.scaling_method in ("pdo_odds", None):
                 round_points = np.rint(points)
+
+                if self.intercept_based:
+                    self.intercept_ = np.rint(self.intercept_)
             elif self.scaling_method == "min_max":
                 round_mip = RoundingMIP()
                 round_mip.build_model(df_scorecard)
@@ -609,6 +628,9 @@ def _fit(self, X, y, sample_weight, metric_special, metric_missing,
                     # Back-up method
                     round_points = np.rint(points)
 
+                if self.intercept_based:
+                    self.intercept_ = np.rint(self.intercept_)
+
             df_scorecard.loc[:, "Points"] = round_points
         self._time_rounding = time.perf_counter() - time_rounding
 
diff --git a/tests/test_scorecard.py b/tests/test_scorecard.py
index 2630dc0..a4aed27 100644
--- a/tests/test_scorecard.py
+++ b/tests/test_scorecard.py
@@ -63,6 +63,26 @@ def test_params():
                               estimator=estimator, intercept_based=1)
         scorecard.fit(X, y)
 
+    with raises(ValueError):
+        scorecard = Scorecard(binning_process=binning_process,
+                              estimator=estimator,
+                              scaling_method=None, rounding=True)
+        scorecard.fit(X, y)
+
+    with raises(ValueError):
+        scorecard = Scorecard(binning_process=binning_process,
+                              estimator=estimator, scaling_method="min_max",
+                              scaling_method_params={'min': 1.1, 'max': 10},
+                              rounding=True)
+        scorecard.fit(X, y)
+
+    with raises(ValueError):
+        scorecard = Scorecard(binning_process=binning_process,
+                              estimator=estimator, scaling_method="min_max",
+                              scaling_method_params={'min': 1, 'max': 10.1},
+                              rounding=True)
+        scorecard.fit(X, y)
+
     with raises(TypeError):
         scorecard = Scorecard(binning_process=binning_process,
                               estimator=estimator, reverse_scorecard=1)
@@ -296,7 +316,7 @@ def test_rounding():
     binning_process = BinningProcess(variable_names)
     estimator = LogisticRegression()
 
-    scaling_method_params = {"min": 200.52, "max": 850.66}
+    scaling_method_params = {"min": 200, "max": 851}
 
     scorecard = Scorecard(binning_process=binning_process,
                           estimator=estimator, scaling_method="min_max",