tslumen.profile.features module¶

Time Series features port over and adaptations of some of the functions originally in R’s TSFeatures package.

See also

tslumen.profile.features.ft_acf(data: pandas.core.series.Series, n_diff: Tuple[int, …] = (0, 1, 2), n_size: Tuple[int, …] = (1, 10)) → pandas.core.series.Series[source]¶: ACF-related features

tslumen.profile.features.ft_adfuller(data: pandas.core.series.Series) → pandas.core.series.Series[source]¶

ADFuller-related features

Parameters: data (pd.Series) – Timeseries data.
Returns: Series with 4 values: adfuller(c), adfuller(ct), adfuller(ctt) and adfuller(nc).
Return type: pd.Series

tslumen.profile.features.ft_cross_pts(data: pandas.core.series.Series) → pandas.core.series.Series[source]¶

Number of times a time series crosses the median.

tslumen.profile.features.ft_entropy(data: pandas.core.series.Series, sampling_frequency: float = 1.0, n_per_segment: Optional[int] = None) → pandas.core.series.Series[source]¶: Calculates spectral entropy of the data and its acf function.

See also

https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.welch.html

tslumen.profile.features.ft_kpss(data: pandas.core.series.Series) → pandas.core.series.Series[source]¶

KPSS-related features

tslumen.profile.features.ft_pacf(data: pandas.core.series.Series, n_diff: Tuple[int, …] = (0, 1, 2), n_size: Tuple[int, …] = (5)) → pandas.core.series.Series[source]¶: PACF-related features

tslumen.profile.features.ft_stl(data: pandas.core.series.Series, freq: Optional[int] = None) → pandas.core.series.Series[source]¶

Calculates features related to STL.

Returns: Series with 4 values: trend, seasonality, acf1(error) and acf10(error).
Return type: pd.Series

tslumen.profile.features.ft_tilewin(data: pandas.core.series.Series) → pandas.core.series.Series[source]¶

Tile window features