Time series - third party Library: tsfresh [feature extraction, feature selection]

One 、tsfresh summary

tsfresh It can automatically calculate a large number of time series characteristics , It contains many feature extraction methods and powerful feature selection algorithms .

tsfresh It is used for system feature engineering from time series and other series data . What these data have in common is that they are sorted by independent variables . The most common independent variable is time （ The time series ）.

There is one named hctsa Of matlab package , It can be used to automatically extract features from time series . It can also be done through pyopy Wrapped in Python Use in hctsa . Other packaging programs available are featuretools、FATS and cesium.

pip install tsfresh

tsfresh Official website :https://tsfresh.readthedocs.io/en/latest/index.html

tsfresh It is used for system feature engineering from time series and other series data . What these data have in common is that they are sorted by independent variables . The most common independent variable is time （ The time series ）.

without tsfresh, You will have to calculate all these characteristics manually ;tsfresh Automatically calculate and automatically return all these features .

Besides ,tsfresh And Python library pandas And compatible scikit-learn.

at present ,tsfresh Not suitable for ：

• For streaming data （ Streaming data refers to data commonly used for online operations , and Time series data is usually used for offline operations ）;
• Train the model on the extracted features （ We don't want to reinvent the wheel , Machine learning model , Please check out Python package scikit-learn）;