A demo code for PU classification proposed by Elkan and Noto 2008
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import numpy as np | |
import matplotlib.pyplot as plt | |
from numpy import random | |
import seaborn as sns | |
from sklearn import metrics | |
from puwrapper import PUWrapper | |
from sklearn.linear_model import LogisticRegression,LogisticRegressionCV | |
sns.set_style("white") | |
random.seed(0) | |
n1=100; n2=500; n=n1+n2; | |
mu1=[0,0]; mu2=[2,2]; Sigma1=0.1*np.identity(2); Sigma2=0.5*np.identity(2); | |
X=np.r_[random.multivariate_normal(mu1,Sigma1,n1), | |
random.multivariate_normal(mu2,Sigma2,n2)] | |
y=np.concatenate([np.repeat(1,n1),np.repeat(0,n2)]) | |
s=np.zeros(n,dtype=np.int32) | |
idxp=np.arange(n)[y==1][:np.int32(n1*0.3)] | |
s[idxp]=1 | |
idx=random.permutation(n); X=X[idx]; y=y[idx]; s=s[idx]; | |
# now s[i] indicates X[i] is positive or unlabeled | |
score=lambda l1,l2: metrics.f1_score(l1,l2,average=None)[1] | |
scorer=metrics.make_scorer(score) | |
from rbfmodel_wrapper import RbfModelWrapper | |
from sklearn.grid_search import GridSearchCV | |
Xce=X[np.random.permutation(len(X))[:100]] | |
base1=GridSearchCV(RbfModelWrapper(LogisticRegression(),Xce=Xce),param_grid={"gamma":np.logspace(-1,1,6)},scoring=scorer) | |
base2=GridSearchCV(RbfModelWrapper(LogisticRegression(),Xce=Xce),param_grid={"gamma":np.logspace(-1,1,6)},scoring=scorer) | |
clf=PUWrapper(base1).fit(X,s) | |
trad_clf=base2.fit(X,s) | |
print "accuracy (PU):",metrics.accuracy_score(y[s==0],clf.predict(X[s==0])) | |
print "pos's F1 (PU):",score(y[s==0],clf.predict(X[s==0])) | |
offset=.5 | |
xx,yy=np.meshgrid(np.linspace(X[:,0].min()-offset,X[:,0].max()+offset,100), | |
np.linspace(X[:,1].min()-offset,X[:,1].max()+offset,100)) | |
label=trad_clf.predict(X) | |
proba=trad_clf.predict_proba(np.c_[xx.ravel(),yy.ravel()]) | |
Z=proba[:,1] | |
Z=Z.reshape(xx.shape) | |
label2=clf.predict(X) | |
proba=clf.predict_proba(np.c_[xx.ravel(),yy.ravel()]) | |
Z2=proba[:,1] | |
Z2=Z2.reshape(xx.shape) | |
""" | |
b1=plt.scatter(X[s==1][:,0],X[s==1][:,1],c="blue",s=50) | |
b2=plt.scatter(X[s==0][:,0],X[s==0][:,1],c="grey",s=50) | |
plt.axis("tight") | |
plt.xlim((X[:,0].min()-offset,X[:,0].max()+offset)) | |
plt.ylim((X[:,1].min()-offset,X[:,1].max()+offset)) | |
plt.legend([b1,b2], | |
["positive","unlabeled"], | |
prop={"size":10},loc="upper left") | |
plt.title("Positive and unlabeled data") | |
plt.tight_layout() | |
plt.show() | |
""" | |
""" | |
b1=plt.scatter(X[y==1][:,0],X[y==1][:,1],c="blue",s=50) | |
b2=plt.scatter(X[y==0][:,0],X[y==0][:,1],c="red",s=50) | |
plt.axis("tight") | |
plt.xlim((X[:,0].min()-offset,X[:,0].max()+offset)) | |
plt.ylim((X[:,1].min()-offset,X[:,1].max()+offset)) | |
plt.legend([b1,b2], | |
["positive","negative"], | |
prop={"size":10},loc="upper left") | |
plt.title("Samples with true label") | |
plt.tight_layout() | |
plt.show() | |
""" | |
""" | |
a1=plt.contour(xx, yy, Z2, levels=[0.5], linewidths=2, colors='green') | |
b1=plt.scatter(X[label==1][:,0],X[label==1][:,1],c="blue",s=50) | |
b2=plt.scatter(X[label==0][:,0],X[label==0][:,1],c="red",s=50) | |
plt.axis("tight") | |
plt.xlim((X[:,0].min()-offset,X[:,0].max()+offset)) | |
plt.ylim((X[:,1].min()-offset,X[:,1].max()+offset)) | |
plt.legend([a1.collections[0],b1,b2], | |
["decision boundary","positive","negative"], | |
prop={"size":10},loc="upper left") | |
plt.title("Result of traditional classification") | |
plt.tight_layout() | |
plt.savefig("result_of_tradclf.png") | |
plt.show() | |
""" | |
a1=plt.contour(xx, yy, Z2, levels=[0.5*clf.c_], linewidths=2, colors='green') | |
b1=plt.scatter(X[label2==1][:,0],X[label2==1][:,1],c="blue",s=50) | |
b2=plt.scatter(X[label2==0][:,0],X[label2==0][:,1],c="red",s=50) | |
plt.axis("tight") | |
plt.xlim((X[:,0].min()-offset,X[:,0].max()+offset)) | |
plt.ylim((X[:,1].min()-offset,X[:,1].max()+offset)) | |
plt.legend([a1.collections[0],b1,b2], | |
["decision boundary","positive","negative"], | |
prop={"size":10},loc="upper left") | |
plt.title("Result of PU classification") | |
plt.tight_layout() | |
plt.show() |
@ankitsethknp It's a local class, not standard to Python nor third-party: https://gist.github.com/nkt1546789/9fbbf2f450779bde60c3
I'm the author of this code.
I switched to this account from nkt1546789
.
Maybe I need to rewrite this code asap.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Please tell me how to install puwrapper in python. I executed the following command in linux-
sudo apt-get install puwrapper
But I got this error-
E: Unable to locate package puwrapper