68 lines
1.9 KiB
Python
68 lines
1.9 KiB
Python
"""
|
|
==============================
|
|
Adaboosted Class Probabilities
|
|
==============================
|
|
|
|
This example fits an Adaboosted decision tree on a classification dataset and
|
|
plots the decision surfaces and class probabilities.
|
|
"""
|
|
print __doc__
|
|
|
|
import pylab as pl
|
|
import numpy as np
|
|
|
|
from sklearn.ensemble import AdaBoostClassifier
|
|
from sklearn.tree import DecisionTreeClassifier
|
|
from sklearn.datasets import make_classification
|
|
|
|
X, y = make_classification(n_samples=1000,
|
|
n_features=2,
|
|
n_classes=2,
|
|
n_informative=2,
|
|
n_redundant=0)
|
|
|
|
bdt = AdaBoostClassifier(DecisionTreeClassifier(min_samples_leaf=100),
|
|
n_estimators=50,
|
|
learn_rate=.5)
|
|
|
|
bdt.fit(X, y)
|
|
|
|
plot_colors = "br"
|
|
plot_step = 0.02
|
|
|
|
pl.figure(figsize=(15, 5))
|
|
|
|
# plot the decision boundaries
|
|
pl.subplot(121)
|
|
x_min, x_max = X[:, 0].min() - 1, X[:, 0].max() + 1
|
|
y_min, y_max = X[:, 1].min() - 1, X[:, 1].max() + 1
|
|
xx, yy = np.meshgrid(np.arange(x_min, x_max, plot_step),
|
|
np.arange(y_min, y_max, plot_step))
|
|
norm = sum(bdt.boost_weights_)
|
|
for weight, tree in zip(bdt.boost_weights_, bdt.estimators_):
|
|
Z = tree.predict(np.c_[xx.ravel(), yy.ravel()])
|
|
Z = Z.reshape(xx.shape)
|
|
cs = pl.contourf(xx, yy, Z, alpha=weight / norm, cmap=pl.cm.Paired)
|
|
pl.axis("tight")
|
|
|
|
# Plot the training points
|
|
for i, c in zip(xrange(2), plot_colors):
|
|
idx = np.where(y == i)
|
|
pl.scatter(X[idx, 0], X[idx, 1], c=c, cmap=pl.cm.Paired)
|
|
pl.axis("tight")
|
|
pl.xlabel("Decision Surfaces")
|
|
|
|
# plot the class probabilities
|
|
pl.subplot(122)
|
|
pl.hist(bdt.predict_proba(X[y==0])[:,-1], bins=20, range=(0, 1),
|
|
facecolor=plot_colors[0],
|
|
label='Class A')
|
|
pl.hist(bdt.predict_proba(X[y==1])[:,-1], bins=20, range=(0, 1),
|
|
facecolor=plot_colors[1],
|
|
label='Class B')
|
|
pl.legend()
|
|
pl.ylabel('Samples')
|
|
pl.xlabel('Class Probability')
|
|
|
|
pl.show()
|