def plot_partial_dependences(model,
X,
var_names,
y=None,
bootstrap_models=None,
pipeline=None,
n_points=250):
"""Convenience function for creating many partial dependency plots."""
fig, axs = plt.subplots(len(var_names), figsize=(12, 3 * len(var_names)))
for ax, name in zip(axs, var_names):
if bootstrap_models:
for M in bootstrap_models[:100]:
print(M)
plot_partial_depenence(ax,
M,
X=X,
var_name=name,
pipeline=pipeline,
alpha=0.8,
linewidth=1,
color="lightblue")
print(model)
plot_partial_depenence(ax,
model,
X=X,
var_name=name,
y=y,
pipeline=pipeline,
color="blue",
linewidth=3)
ax.set_title("{} Partial Dependence".format(name))
return fig, axs
def plot_partial_dependences(model,
X,
var_names,
y=None,
bootstrap_models=None,
pipeline=None,
n_points=250):
"""Creates many partial dependency plots.
INPUT:
model:
A sklearn model
Call fit() on model before using plot_feature_importances.
X:
A dataframe or array containing independent variables
that fit in pipeline.
var_names:
The column names of the x variables in the
dataframe you want plotted.
y:
A dataframe or array, the dependent variable.
Plotted in grey.
pipeline:
Runs on df_X (without y), already fit to df_X.
n_points:
The number of points to plot.
*kwargs:
Extra arguments passed to plot_partial_dependence.
OUTPUT:
an array of partial dependence plots, describing each varible's
contibution to the regression.
"""
fig, axs = plt.subplots(len(var_names), figsize=(12, 3 * len(var_names)))
for ax, name in zip(axs, var_names):
if bootstrap_models:
for M in bootstrap_models[:10]:
# print(M)
plot_partial_depenence(ax,
M,
X=X,
var_name=name,
pipeline=pipeline,
alpha=0.8,
linewidth=1,
color="lightblue")
# print(model)
plot_partial_depenence(ax,
model,
X=X,
var_name=name,
y=y,
pipeline=pipeline,
color="blue",
linewidth=3)
ax.set_title("{} Partial Dependence".format(name))
return fig, axs
def shaped_plot_partial_dependences(model,
df,
y_var_name,
pipeline=None,
n_points=250,
**kwargs):
X_features = list(df.columns)
X_features.remove(y_var_name)
if len(X_features) > 1:
num_plot_rows = int(np.ceil(len(X_features) / 2.0))
fig, axs = plt.subplots(num_plot_rows,
2,
figsize=(14, 3 * num_plot_rows))
for i, X_feature in enumerate(X_features):
# print(model)
plot_partial_depenence(axs.flatten()[i],
model=model,
X=df.drop(y_var_name, axis=1),
var_name=X_feature,
y=df[y_var_name],
pipeline=pipeline,
n_points=n_points,
**kwargs)
axs.flatten()[i].set_title("{}: Partial Dependence Plot {}".format(
X_feature, model.__class__.__name__))
elif len(X_features) == 1:
fig, axs = plt.subplots(len(X_features),
1,
figsize=(14, 4.5 * len(X_features)))
for i, X_feature in enumerate(X_features):
plot_partial_depenence(axs,
model=model,
X=df.drop(y_var_name, axis=1),
var_name=X_feature,
y=df[y_var_name],
pipeline=pipeline,
n_points=n_points,
**kwargs)
axs.set_title("{}: Partial Dependence Plots {}".format(
X_feature, model.__class__.__name__))
fig.set_title("Partial Dependence Plots for " +
model.__class__.__name__)
# fig.set_tight_layout(tight = True) #this doesn't work!!!
fig.tight_layout(
pad=2) # 'tight_layout' must be used in calling script as well
else:
print('No Features to Plot')
def shaped_plot_partial_dependences(model,
df,
y_var_name,
pipeline=None,
n_points=250,
**kwargs):
"""Creates many partial dependency plots.
INPUT:
model:
A sklearn model
Call fit() on model before using plot_feature_importances.
df:
A dataframe containing independent variables y
and dependent variables X.
y_var_name:
String, the column name of the dependent y variable in the dataframe
pipeline:
Runs on df_X (without y), already fit to df_X.
n_points:
the number of points to plot
*kwargs:
extra arguments passed to plot_partial_dependence
OUTPUT:
an array of partial dependence plots, describing each varible's contibution to the regression.
"""
X_features = list(df.columns)
X_features.remove(y_var_name)
if len(X_features) > 1:
num_plot_rows = int(np.ceil(len(X_features) / 2.0))
fig, axs = plt.subplots(num_plot_rows,
2,
figsize=(14, 3 * num_plot_rows))
for i, X_feature in enumerate(X_features):
# print(model)
plot_partial_depenence(axs.flatten()[i],
model=model,
X=df.drop(y_var_name, axis=1),
var_name=X_feature,
y=df[y_var_name],
pipeline=pipeline,
n_points=n_points,
**kwargs)
axs.flatten()[i].set_title("{}: Partial Dependence Plot {}".format(
X_feature, model.__class__.__name__))
elif len(X_features) == 1:
fig, axs = plt.subplots(len(X_features),
1,
figsize=(14, 4.5 * len(X_features)))
for i, X_feature in enumerate(X_features):
plot_partial_depenence(axs,
model=model,
X=df.drop(y_var_name, axis=1),
var_name=X_feature,
y=df[y_var_name],
pipeline=pipeline,
n_points=n_points,
**kwargs)
axs.set_title("{}: Partial Dependence Plots {}".format(
X_feature, model.__class__.__name__))
fig.set_title("Partial Dependence Plots for " +
model.__class__.__name__)
# fig.set_tight_layout(tight = True) #this doesn't work!!!
# 'tight_layout' must be used in calling script as well
fig.tight_layout(pad=2)
else:
print('No Features to Plot')