---
id: tbats
version: "1.1.3"
license: MIT License
license_treatment: permissive
maintenance: abandoned
---

# tbats — BATS and TBATS for time series forecasting

License: permissive · Maintenance: abandoned · Downloads: 710.6K/mo

## What it is and what it does

TBATS implements Bayesian Automatic Time Series forecasting methods designed to handle time series with multiple seasonal patterns and trend components. It combines exponential smoothing with automatic model selection, testing various configurations (including Box-Cox transformation and ARMA error modeling) to find the best fit. The package wraps numpy, scipy, scikit-learn, and pmdarima to perform this model search and fitting.

You provide a time series and optionally specify seasonal periods; TBATS fits a model, then forecasts forward. It exposes fitted values, residuals, model parameters, and summary statistics. The package is a Python port of the R forecast package's BATS/TBATS implementation. Fitting can be computationally intensive on long series, and the documentation includes troubleshooting guidance for slow convergence and multiprocessing issues.

Use it for:

- Forecast retail sales or demand with weekly and yearly seasonal cycles without manual model specification.
- Analyze electricity load or utility consumption with multiple overlapping seasonal patterns.
- Predict web traffic or system metrics that exhibit both daily and weekly seasonality.
- Generate point forecasts and residual diagnostics for time series with complex, non-standard seasonal structures.
- Prototype time series forecasting pipelines where automatic model selection is preferred over manual tuning.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Implements BATS and TBATS exponential smoothing methods for forecasting time series with complex seasonal patterns, built on numpy, scipy, scikit-learn, and pmdarima.

Yes, if you need BATS/TBATS forecasting and can tolerate an abandoned package. The library is stable, has low install friction, carries no known vulnerabilities, and is permissively licensed. However, the last release was 2023-04-17 with no active maintenance. Use it for prototyping or production work where you can accept no further updates; avoid it if you need ongoing support or compatibility with future dependency versions.

## Install

pip install tbats
uv add tbats
poetry add tbats

## Installing tbats

Before you install: Low friction: pure Python wheel with four well-established scientific dependencies. However, the package is abandoned—last release was 2023-04-17 and no commits since. No active maintenance or security updates.

License in practice: MIT License (permissive): you can use, modify, and distribute this package freely with minimal restrictions, though you must retain the license notice.

Quickstart:

pip install tbats

from tbats import TBATS
import numpy as np

if __name__ == '__main__':
    y = np.array([...])  # your time series
    estimator = TBATS(seasonal_periods=[14, 30.5])
    fitted_model = estimator.fit(y)
    forecast = fitted_model.forecast(steps=14)

On Windows, fitting requires wrapping code in `if __name__ == '__main__':` due to multiprocessing behavior. Fitting can be slow on long time series; consider disabling Box-Cox transformation and ARMA errors or setting n_jobs=1 if parallelization freezes.

Verify before relying:

- Whether the package remains compatible with current versions of numpy, scipy, scikit-learn, and pmdarima given the days since last release.
- Whether the R forecast package equivalence (stated in docs) is maintained across recent updates to the R implementation.

## Package facts

- License: MIT License (permissive)
- Python support: unspecified
- Install friction: low
- Maintenance: abandoned
- Downloads: 710.6K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

time series forecasting seasonal, BATS TBATS exponential smoothing, complex seasonal pattern forecasting, time series with multiple seasonality, automated forecasting methods, seasonal decomposition forecasting, time-series-forecasting, seasonal-decomposition

[View on SkillFed](https://skillfed.io/packages/tbats) · [View on PyPI](https://pypi.org/project/tbats/)