seasonal 0.3.0

Estimate trend and seasonal effects in a timeseries

Robustly estimate and remove trend and periodicity in a timeseries.

Seasonal can recover sharp trend and period estimates from noisy timeseries data with only a few periods. It is intended for estimating season, trend, and level when initializing structural timeseries models like Holt-Winters. Input samples are assumed evenly-spaced from a continuous-time signal with noise but no anomalies.

In this package, trend removal is in service of isolating and estimating periodic (non-trend) variation. “trend” is in the sense of Cleveland’s STL decomposition – a lowpass smoothing of the data, rather than a single linear trend (though you can opt for this). Detrending is accomplishd by a coarse fitted spline or a median filter.

The seasonal estimate will be a list of period-over-period averages at each seasonal offset. You may specify a period length, or have it estimated from the data. The latter is an interesting capability of this package.

See README.md for details on installation, API, theory, and examples.

Dependencies

package: numpy, scipy extras: pandas, matplotlib