CODEWITHSUNDEEP
pythonscikit-learnfeature-selection
Anuradha
Anuradha

Reputation: 580

Feature Selection

I tried to do recursive feature selection in scikit learn with following code.

from sklearn import datasets, svm
from sklearn.feature_selection import SelectKBest, f_classif
from sklearn.feature_selection import RFE
import numpy as np

input_file_iris = "/home/anuradha/Project/NSL_KDD_master/Modified/iris.csv"
dataset = np.loadtxt(input_file_iris, delimiter=",")
X = dataset[:,0:4]
y = dataset[:,4]

estimator= svm.OneClassSVM(nu=0.1, kernel="rbf", gamma=0.1)

selector = RFE(estimator,3, step=1)
selector = selector.fit(X,y)

But it gives following error

Traceback (most recent call last):
File "/home/anuradha/PycharmProjects/LearnPython/Scikit-learn/univariate.py", line 30, in <module>
File "/usr/local/lib/python2.7/dist-packages/sklearn/feature_selection/rfe.py", line 131, in fit
return self._fit(X, y)


File "/usr/local/lib/python2.7/dist-packages/sklearn/feature_selection/rfe.py", line 182, in _fit



raise RuntimeError('The classifier does not expose '
RuntimeError: The classifier does not expose "coef_" or 
"feature_importances_" attributes

Please some one can help me to solve this or guide me to another solution

Upvotes: 2

Views: 1015

Answers (1)

MhFarahani
MhFarahani

Reputation: 970

Change your kernel to linear and your code would work.

Besides, svm.OneClassSVM is used for unsupervised outlier detection. Are you sure that you want to use it as estimator? Or perhaps you want to use svm.SVC(). Look the following link for documentation.

http://scikit-learn.org/stable/modules/generated/sklearn.svm.OneClassSVM.html

Lastly, iris data set is already available in sklearn. You have imported the sklearn.datasets. So you can simply load iris as:

iris = datasets.load_iris()
X = iris.data
y = iris.target

Upvotes: 2

Related Questions