statsforecast
High-performance statistical and econometric time series forecasting models for Python environments.
Accelerated AutoARIMA, ETS, and MSTL models scaling to millions of time series.
StatsForecast delivers high-speed implementations of traditional statistical forecasting models like AutoARIMA and ETS in Python. It provides a faster alternative to libraries like Prophet or pmdarima for high-volume demand and trend prediction. You should use it if you need scalable, classic statistical time series modeling.
What it is
The library features optimized implementations of algorithms like AutoARIMA, AutoETS, MSTL, and Theta with standard scikit-learn style methods. It supports probabilistic forecasting, prediction intervals, exogenous regressors, and anomaly detection. You can distribute fitting across millions of series using integrations with Ray, Spark, and Dask.
Who it's for
It suits data scientists and quantitative analysts who need fast statistical baselines or production time series models in Python environments. It is not suitable for non-programmers or teams looking for a visual no-code analytics interface.
Key features
Accelerated Statistical Models
Provides high-speed implementations of AutoARIMA, AutoETS, AutoCES, MSTL, and Theta.
Distributed Scale Compatibility
Executes natively across distributed computing frameworks like Spark, Dask, and Ray.
Exogenous Variable Support
Incorporates external regressors such as weather or prices directly into statistical models.
Probabilistic Prediction Intervals
Generates point forecasts along with probabilistic intervals for risk quantification.
In-sample Anomaly Detection
Identifies anomalies across time series signals using in-sample prediction intervals.
Up and running in 3 steps
Our honest verdict
What's great
- Significantly faster execution compared to pmdarima, Prophet, and statsmodels.
- Includes familiar scikit-learn fit and predict syntax for straightforward pipeline integration.
- Natively scales across distributed clusters using Ray, Dask, or Spark to process millions of series.
- Offers built-in support for multiple seasonalities, exogenous features, and prediction intervals.
Where it falls short
- Requires custom compute infrastructure or cloud cluster resources when processing massive datasets with distributed backends.
- Lacks a graphical user interface, requiring proficient Python programming skills to deploy.
- Focuses primarily on univariate statistical methods rather than complex deep learning architectures.
Pricing & license
StatsForecast is free and open source under the Apache-2.0 license. The only direct cost is the compute power or server infrastructure you supply to run your workloads.
Open remote AI-training roles
874 open remote AI-training roles, updated November 15, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
How does StatsForecast handle massive scale?
It integrates with distributed backends like Ray, Spark, and Dask to run models across millions of series concurrently.
Can I include external features like weather or promotions?
Yes, models like AutoARIMA and Theta allow you to pass exogenous regressors alongside target time series.
