4

I am new to machine learning and am trying to understand cross_val_score uses Kfold to split the data to k folds.

kf = KFold(n_splits=2)
cv_results =cross_val_score(model, X_train, Y_train, cv=kf)

I know kfold splits the data but I tried printing it out

dataset = [[1,1,1],[2,2,2],[3,3,3],[4,4,4],[5,5,5],[6,6,6],[7,7,7],[8,8,8]]
kf =  KFold(n_splits=2)
print kf

>>> KFold(n_splits=2, random_state=None, shuffle=False)

It doesn't show the k folds but then how does cross_val_score get all the folds?

Venkatachalam
  • 16,288
  • 9
  • 49
  • 77
TheLastCoder
  • 610
  • 2
  • 6
  • 15

2 Answers2

4

You need to call Kf.split(dataset) to actually split the data. Click here to see how KFold works

Just to make it clear, KFold is a class and not a function.

kf = KFold(n_splits=2) creates an object of KFold. and print kf will just print out the class object.

and when you callcross_val_score(model, X_train, Y_train, cv=kf) you are passing the object kf to cross_val_score function where kf.split(X_train) would be called to split X_train into 2 folds. Y_train would also be splitted similarly.

DataCruncher
  • 850
  • 7
  • 11
0

Try this

kf = KFold(n_splits=2)
generator = kf.split(dataset)
for train, test in generator:
    print "*" * 20
    print "Training Data:"
    for i in train:
        print dataset[i]
    print "Test Data:"
    for j in test:
        print dataset[j]

kf.split(dataset) returns a generator. Iterating through the generator would give you all the folds

danny ruben
  • 21
  • 5
  • We are not passing the k folds to the cross_val_score function. Just the KFold object. I mean, this does help if you want to see all the folds created by KFold.split() function – DataCruncher May 25 '19 at 17:07
  • Thanks, I got the K folds, now how do I pass it to cross_val_score? – TheLastCoder May 25 '19 at 17:09