tsfresh

What it is

An open-source Python package for systematic time-series feature extraction based on scalable hypothesis tests. It automatically computes hundreds of features, from basic statistics like peak counts and averages to measures such as time reversal symmetry, and applies a built-in multiple-test filtering procedure to remove features with low explanatory power for a given regression or classification task. Quant researchers and data scientists who want to automate feature engineering on sampled data or event sequences will find it useful. Because most extracted features may be irrelevant to a particular task, users should rely on the filtering step rather than assume every feature is informative.