Compatibility with Scikit-Learn
The classes provided in this package (i.e., GapLeavePOut,
GapKFold, and GapRollForward) are rarely used directly.
More frequently, their instances are sent to a scikit-learn
cross-validator. This page shows this usage.
Let us start by preparing the environment:
import numpy as np
from sklearn import datasets
from sklearn import svm
from sklearn.model_selection import cross_val_score
from tscv import GapKFold
In this code snippet, sklearn.model_selection.cross_val_score
is a cross-validator provided by scikit-learn, and GapKFold,
provided by this package, is referred to as splitter in this page.
Now let us initiate the Iris dataset and the SVM algorithm:
iris = datasets.load_iris()
clf = svm.SVC(kernel='linear', C=1)
Note
The Iris dataset is not a time series. Here we use it just to showcase the usage of the package.
Now instantiate a splitter:
cv = GapKFold(n_splits=5, gap_before=5, gap_after=5)
and send it to the cross-validator along with the dataset and the algorithm:
scores = cross_val_score(clf, iris.data, iris.target, cv=cv)
scores saves the cross-validation score and is used as the
criterion to evaluate algorithms:
>>> scores
array([1. , 1. , 0.83333333, 0.96666667, 0.83333333])