A demo code for ranking SVM.

ranking_svm_demo.py

"""
A demo code for ranking SVM
The data used in this code comes from http://download.joachims.org/svm_light/examples/example3.tar.gz
"""

import numpy as np
import itertools
from sklearn.linear_model import SGDClassifier
from sklearn.grid_search import GridSearchCV
np.random.seed(0)

def transform(X,y):
    order=np.argsort(y)
    Xordered=X[order]
    yordered=y[order]
    Xnew=[]
    ynew=[]
    for i,j in itertools.combinations(range(X.shape[0]),2):
        Xnew.append(Xordered[i]-Xordered[j])
        ynew.append(1)
    return Xnew,ynew

def transform_randomly(X,y):
    idx=np.random.permutation(len(y))
    X=X[idx]
    y=y[idx]
    Xnew=[]
    ynew=[]
    for i,j in itertools.combinations(range(X.shape[0]),2):
        Xnew.append(X[i]-X[j])
        if y[i]>=y[j]:
            ynew.append(1)
        else:
            ynew.append(-1)
    return Xnew,ynew

Xtr=np.array([[1, 1, 0, 0.2, 0],
              [0, 0, 1, 0.1, 1],
              [0, 1, 0, 0.4, 0],
              [0, 0, 1, 0.3, 0],
              [0, 0, 1, 0.2, 0],
              [1, 0, 1, 0.4, 0],
              [0, 0, 1, 0.1, 0],
              [0, 0, 1, 0.2, 0],
              [0, 0, 1, 0.1, 1],
              [1, 1, 0, 0.3, 0],
              [1, 0, 0, 0.4, 1],
              [0, 1, 1, 0.5, 0]])
ytr=np.array([3,2,1,1,1,2,1,1,2,3,4,1])
qids=np.array([1,1,1,1,2,2,2,2,3,3,3,3])

Xte=np.array([[1, 0, 0, 0.2, 1],
              [1, 1, 0, 0.3, 0],
              [0, 0, 0, 0.2, 1],
              [0, 0, 1, 0.2, 0]])
yte=np.array([4,3,2,1])

Xtr_transformed=[]
ytr_transformed=[]
for qid in np.unique(qids):
    #Xp,yp=transform(Xtr[qids==qid],ytr[qids==qid])
    Xp,yp=transform_randomly(Xtr[qids==qid],ytr[qids==qid])
    Xtr_transformed+=Xp
    ytr_transformed+=yp

Xtr_transformed=np.array(Xtr_transformed)
ytr_transformed=np.array(ytr_transformed)

clf=SGDClassifier(loss="hinge",penalty="l2")
clf=GridSearchCV(clf,param_grid={"alpha":np.logspace(-1,7,10)}).fit(Xtr_transformed,ytr_transformed).best_estimator_

scores=clf.decision_function(Xte).ravel()
print "true ranking:",np.argsort(yte)[::-1]
print "predicted ranking:",np.argsort(scores)[::-1]
print "scores:",scores