add Nina's improved plotting functions

Author: JohnMount

README.txt

@@ -1,7 +1,7 @@
 Win Vector LLC extras for teaching data science in Python 3
 
 pip install wvpy
-# or: pip install https://github.com/WinVector/wvpy/raw/master/dist/wvpy-0.1.2.tar.gz
+# or: pip install https://github.com/WinVector/wvpy/raw/master/dist/wvpy-0.1.3.tar.gz
 
 ~/anaconda3/bin/python3
 

clean.bash

@@ -0,0 +1,8 @@
+
+pushd pkg
+rm -rf dist build wvpy.egg-info wvpy/__pycache__ tests/__pycache__
+popd
+pip uninstall -y wvpy
+
+
+

pkg/build/lib/wvpy/util.py

@@ -93,7 +93,7 @@ def plot_roc(prediction, istrue, title="Receiver operating characteristic plot")
     return auc
 
 
-def dual_density_plot(probs, istrue):
+def dual_density_plot(probs, istrue, title="Double density plot"):
     """plot a dual density plot of numeric prediction probs against boolean istrue"""
     matplotlib.pyplot.gcf().clear()
     preds_on_positive = [probs[i] for i in range(len(probs)) if istrue[i]]
@@ -102,6 +102,7 @@ def dual_density_plot(probs, istrue):
     seaborn.kdeplot(preds_on_negative, label="negative examples", shade=True)
     matplotlib.pyplot.ylabel("density of examples")
     matplotlib.pyplot.xlabel("model score")
+    matplotlib.pyplot.title(title)
     matplotlib.pyplot.show()
 
 
@@ -129,9 +130,11 @@ def dual_hist_plot_proba1(probs, istrue):
     matplotlib.pyplot.show()
 
 
-def gain_curve_plot(prediction, outcome):
+def gain_curve_plot(prediction, outcome, title="Gain curve plot"):
     """plot cumulative outcome as a function of prediction order (descending)"""
     df = pandas.DataFrame({"prediction": prediction, "outcome": outcome})
+
+    # compute the gain curve
     df.sort_values(["prediction"], ascending=[False], inplace=True)
     df["fraction_of_observations_by_prediction"] = [
         (1 + i) / df.shape[0] for i in range(df.shape[0])
@@ -140,15 +143,37 @@ def gain_curve_plot(prediction, outcome):
     df["cumulative_outcome_fraction"] = df["cumulative_outcome"] / numpy.max(
         df["cumulative_outcome"]
     )
-    seaborn.scatterplot(
+
+    # compute the wizard curve
+    df.sort_values(["outcome"], ascending=[False], inplace=True)
+    df["fraction_of_observations_by_wizard"] = [
+        (1 + i) / df.shape[0] for i in range(df.shape[0])
+    ]
+    df["cumulative_outcome_by_wizard"] = df["outcome"].cumsum()
+    df["cumulative_outcome_fraction_wizard"] = df["cumulative_outcome_by_wizard"] / numpy.max(
+        df["cumulative_outcome_by_wizard"]
+    )
+
+    seaborn.lineplot(
         x="fraction_of_observations_by_prediction",
         y="cumulative_outcome_fraction",
         data=df,
     )
+
+    seaborn.lineplot(
+        x="fraction_of_observations_by_wizard",
+        y="cumulative_outcome_fraction_wizard",
+        color="gray",
+        linestyle="--",
+        data=df,
+    )
+
     seaborn.lineplot(x=[0, 1], y=[0, 1], color="red")
+    matplotlib.pyplot.title(title)
+    matplotlib.pyplot.show()
 
 
-def lift_curve_plot(prediction, outcome):
+def lift_curve_plot(prediction, outcome, title="Lift curve plot"):
     """plot lift as a function of prediction order (descending)"""
     df = pandas.DataFrame({"prediction": prediction, "outcome": outcome})
     df.sort_values(["prediction"], ascending=[False], inplace=True)
@@ -162,8 +187,10 @@ def lift_curve_plot(prediction, outcome):
     df["lift"] = (
         df["cumulative_outcome_fraction"] / df["fraction_of_observations_by_prediction"]
     )
-    seaborn.scatterplot(x="fraction_of_observations_by_prediction", y="lift", data=df)
+    seaborn.lineplot(x="fraction_of_observations_by_prediction", y="lift", data=df)
     matplotlib.pyplot.axhline(y=1, color="red")
+    matplotlib.pyplot.title(title)
+    matplotlib.pyplot.show()
 
 
 def dual_hist_plot(probs, istrue):

pkg/dist/wvpy-0.1.2-py3-none-any.whl

@@ -7,7 +7,7 @@
 """
 
 setuptools.setup(name='wvpy',
-      version='0.1.2',
+      version='0.1.3',
       author='John Mount',
       author_email='[email protected]',
       url='https://github.com/WinVector/wvpy',

pkg/dist/wvpy-0.1.2.tar.gz

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: wvpy
-Version: 0.1.2
+Version: 0.1.3
 Summary: Simple utilities for teaching Pandas and scikit learn.
 Home-page: https://github.com/WinVector/wvpy
 Author: John Mount

pkg/dist/wvpy-0.1.3-py3-none-any.whl

@@ -93,7 +93,7 @@ def plot_roc(prediction, istrue, title="Receiver operating characteristic plot")
     return auc
 
 
-def dual_density_plot(probs, istrue):
+def dual_density_plot(probs, istrue, title="Double density plot"):
     """plot a dual density plot of numeric prediction probs against boolean istrue"""
     matplotlib.pyplot.gcf().clear()
     preds_on_positive = [probs[i] for i in range(len(probs)) if istrue[i]]
@@ -102,6 +102,7 @@ def dual_density_plot(probs, istrue):
     seaborn.kdeplot(preds_on_negative, label="negative examples", shade=True)
     matplotlib.pyplot.ylabel("density of examples")
     matplotlib.pyplot.xlabel("model score")
+    matplotlib.pyplot.title(title)
     matplotlib.pyplot.show()
 
 
@@ -129,9 +130,11 @@ def dual_hist_plot_proba1(probs, istrue):
     matplotlib.pyplot.show()
 
 
-def gain_curve_plot(prediction, outcome):
+def gain_curve_plot(prediction, outcome, title="Gain curve plot"):
     """plot cumulative outcome as a function of prediction order (descending)"""
     df = pandas.DataFrame({"prediction": prediction, "outcome": outcome})
+
+    # compute the gain curve
     df.sort_values(["prediction"], ascending=[False], inplace=True)
     df["fraction_of_observations_by_prediction"] = [
         (1 + i) / df.shape[0] for i in range(df.shape[0])
@@ -140,15 +143,37 @@ def gain_curve_plot(prediction, outcome):
     df["cumulative_outcome_fraction"] = df["cumulative_outcome"] / numpy.max(
         df["cumulative_outcome"]
     )
-    seaborn.scatterplot(
+
+    # compute the wizard curve
+    df.sort_values(["outcome"], ascending=[False], inplace=True)
+    df["fraction_of_observations_by_wizard"] = [
+        (1 + i) / df.shape[0] for i in range(df.shape[0])
+    ]
+    df["cumulative_outcome_by_wizard"] = df["outcome"].cumsum()
+    df["cumulative_outcome_fraction_wizard"] = df["cumulative_outcome_by_wizard"] / numpy.max(
+        df["cumulative_outcome_by_wizard"]
+    )
+
+    seaborn.lineplot(
         x="fraction_of_observations_by_prediction",
         y="cumulative_outcome_fraction",
         data=df,
     )
+
+    seaborn.lineplot(
+        x="fraction_of_observations_by_wizard",
+        y="cumulative_outcome_fraction_wizard",
+        color="gray",
+        linestyle="--",
+        data=df,
+    )
+
     seaborn.lineplot(x=[0, 1], y=[0, 1], color="red")
+    matplotlib.pyplot.title(title)
+    matplotlib.pyplot.show()
 
 
-def lift_curve_plot(prediction, outcome):
+def lift_curve_plot(prediction, outcome, title="Lift curve plot"):
     """plot lift as a function of prediction order (descending)"""
     df = pandas.DataFrame({"prediction": prediction, "outcome": outcome})
     df.sort_values(["prediction"], ascending=[False], inplace=True)
@@ -162,8 +187,10 @@ def lift_curve_plot(prediction, outcome):
     df["lift"] = (
         df["cumulative_outcome_fraction"] / df["fraction_of_observations_by_prediction"]
     )
-    seaborn.scatterplot(x="fraction_of_observations_by_prediction", y="lift", data=df)
+    seaborn.lineplot(x="fraction_of_observations_by_prediction", y="lift", data=df)
     matplotlib.pyplot.axhline(y=1, color="red")
+    matplotlib.pyplot.title(title)
+    matplotlib.pyplot.show()
 
 
 def dual_hist_plot(probs, istrue):