Omnicast

Automatic statistical forecasting for Python with one consistent, interval-aware API.

Status

v0.1 alpha. The API is usable, but model coverage and R parity fixtures are still growing.

Every estimator in this library follows the same shape:

model = SomeForecaster(...)
model.fit(y)
forecast = model.predict(horizon=6, level=[80, 95])

y is a pandas.Series (or anything coercible to one) indexed by a PeriodIndex, a DatetimeIndex with a regular frequency, a RangeIndex, or a plain numeric index. forecast is a ForecastResult carrying the point forecast and every requested prediction interval.

Contents

Getting started

• Installation
  • Local development
• Quickstart
  • The pieces
  • Choosing a model
• Evaluation
  • Backtesting
  • Metrics

Model guide & examples

• Model guide & examples
  • NaiveForecaster
  • SeasonalNaiveForecaster
  • MeanForecaster
  • DriftForecaster
  • ThetaForecaster
  • ETSForecaster
  • ARIMAForecaster
  • AutoARIMAForecaster
  • LSTMForecaster
  • AutoForecaster
  • Real-data walkthrough

API reference

• API reference
  • Base estimator
  • Models
  • AutoForecaster
  • Evaluation
  • Metrics
  • Results
  • Plotting