Automatically extract hundreds of relevant features to solve your time series problem with ease
from tsfresh import extract_relevant_features
from tsfresh.examples.robot_execution_failures import load_robot_execution_failures
timeseries, y = load_robot_execution_failures()
features = extract_relevant_features(timeseries, y, column_id="id", column_sort="time")
Use the extracted relevant features to train your usual ML model to distinguish between different time series classes.
No need for complicated methods! With
tsfresh your time series forecasting problem becomes a usual regression problem.
Detect interesting patterns and outliers in your time series data by clustering the extracted features or training an ML method on them.
tsfresh is the basis for your next time series project!
tsfresh extracts features on your time series data simple and fast, so you can spend more time on using these features.
The feature library in tsfresh contains features calculators from multiple domains, so you can get the best out of your data
Follow our QuickStart tutorial and set up your first feature extraction project on time series.
Read through the documentation on how the feature selection and all the other algorithms work.
Find out, how to apply tsfresh on large data samples using multiprocessing, dask or spark.
$ pip install -U tsfresh
$ conda install -c conda-forge tsfresh