Scikit-learn cross val score: too many indices for array

Question

I have the following code

 from sklearn.ensemble import ExtraTreesClassifier
 from sklearn.cross_validation import cross_val_score
 #split the dataset for tr         


        

Answer 1

It seems to be fixable if you specify the target labels as a single data column from Pandas. If the target has multiple columns, I get a similar error. For example try:

labels = train['Y']

Answer 2

try target:

y=df['Survived'] 

instead , i used

y=df[['Survived']] 

which made the target y a dateframe, it seems series would be ok

Answer 3

Adding .ravel() to the Y/Labels variable passed into the formula helped solve this problem within KNN as well.

Answer 4

You might need to play with the dimensions a bit, e.g.

et_score = cross_val_score(et, features, labels, n_jobs=-1)[:,n]

or

 et_score = cross_val_score(et, features, labels, n_jobs=-1)[n,:]

n being the dimension.

Answer 5

When we do cross validation in scikit-learn, the process requires an (R,) shape label instead of (R,1). Although they are the same thing to some extend, their indexing mechanisms are different. So in your case, just add:

c, r = labels.shape
labels = labels.reshape(c,)

before passing it to the cross-validation function.