Time series interpolating with aeon

Suppose we have a set of time series with different lengths, i.e. different number of time points. Currently, most of aeon’s functionality requires equal-length time series, so to use aeon, we need to first converted our data into equal-length time series. In this tutorial, you will learn how to use the TSInterpolator to do so.

from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline

from aeon.classification.convolution_based import RocketClassifier
from aeon.datasets import load_basic_motions
from aeon.transformations.compose import ColumnConcatenator

Ordinary situation

Here is a normal situation, when all time series have same length. We load an example data set from aeon and train a classifier.

X, y = load_basic_motions(return_type="nested_univ")
X_train, X_test, y_train, y_test = train_test_split(X, y)

steps = [
    ("concatenate", ColumnConcatenator()),
    ("classify", RocketClassifier()),
]
clf = Pipeline(steps)
clf.fit(X_train, y_train)
clf.score(X_test, y_test)

0.25

If time series are unequal length, aeon’s algorithm may raise an error

Now we are going to spoil the data set a little bit by randomly cutting the time series. This leads to unequal-length time series. Consequently, we have an error while attempt to train a classifier.

from aeon.datasets import load_plaid

X, y = load_plaid(return_type="nested_univ")
X_train, X_test, y_train, y_test = train_test_split(X, y)

try:
    clf = RocketClassifier()
    clf = Pipeline(steps)
    clf.fit(X_train, y_train)
    clf.score(X_test, y_test)
except ValueError as e:
    print(f"ValueError: {e}")

IndexError: Data seen by RocketClassifier instance has unequal length series, but this RocketClassifier instance cannot handle unequal length series. Calls with unequal length series may result in error or unreliable results.

Now the interpolator enters

Now we use our interpolator to resize time series of different lengths to user-defined length. Internally, it uses linear interpolation from scipy and draws equidistant samples on the user-defined number of points.

After interpolating the data, the classifier works again.

from aeon.transformations.collection.interpolate import TSInterpolator

steps = [
    ("transform", TSInterpolator(50)),
    ("classify", RocketClassifier()),
]
clf = Pipeline(steps)
clf.fit(X_train, y_train)
clf.score(X_test, y_test)

0.7732342007434945

---

Generated using nbsphinx. The Jupyter notebook can be found here.