嗨,我正在尝试将VotingClassifier与我的GradientBoostingClassifier一起使用,我将其包装起来以便利用sample_weight。
但是,我收到以下错误,无法弄清楚如何解决。

代码:

class MyGradientBoostingClassifier(GradientBoostingClassifier):
    def fit(self, X , y=None):
        return super(GradientBoostingClassifier, self).fit(X, y, sample_weight=y)


rf =  RandomForestClassifier(n_jobs=-1)
mygb = MyGradientBoostingClassifier()

vc = VotingClassifier(estimators=[('rf', rf), ('mygb', mygb)],
                        voting='soft',
                        weights=[1,2])

mygb.fit(X5, y5)


y的样例是[ 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 1. 0. 0. 0. 0.],它是np数组

错误:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-62-c56d4cac146f> in <module>()
     13                         weights=[1,2])
     14
---> 15 mygb.fit(X5, y5)

<ipython-input-62-c56d4cac146f> in fit(self, X, y)
      3         print np.shape(y), np.shape(X), Counter(y), type(y)
      4         print y[:20]
----> 5         return super(GradientBoostingClassifier, self).fit(X, y, sample_weight=y)
      6
      7

/Users/a/anaconda/lib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyc in fit(self, X, y, sample_weight, monitor)
    987
    988             # fit initial model - FIXME make sample_weight optional
--> 989             self.init_.fit(X, y, sample_weight)
    990
    991             # init predictions

/Users/a/anaconda/lib/python2.7/site-packages/sklearn/ensemble/gradient_boosting.pyc in fit(self, X, y, sample_weight)
    117
    118         if neg == 0 or pos == 0:
--> 119             raise ValueError('y contains non binary labels.')
    120         self.prior = self.scale * np.log(pos / neg)
    121

ValueError: y contains non binary labels.

最佳答案

对于分类模型,y应该是整数类标签(0和1),因此将其用作分类目标和样本权重都没有意义。

该模型将忽略权重为0的所有样本,并且不可能仅使用来自训练集中同一类的样本来训练二进制分类模型。

09-07 10:29