StatsForecast is a Python library for time series forecasting that implements statistical and econometric models with a focus on speed and scalability.
The library addresses the problem that existing Python alternatives for statistical forecasting models are slow, inaccurate, and do not scale well. StatsForecast provides optimized implementations of widely used univariate forecasting models including automatic ARIMA, ETS, CES, and Theta, along with a battery of benchmarking models. It supports probabilistic forecasting with confidence intervals, exogenous variables, static covariates, and anomaly detection. The tool uses familiar sklearn-style syntax with fit and predict methods, and integrates natively with distributed computing frameworks including Spark, Dask, and Ray.
Teams should choose this tool if they need to forecast at scale or require statistical models as production baselines. It suits projects handling large numbers of time series, from hundreds to millions, where performance and accuracy matter. The README indicates the tool can replace Facebook's Prophet with minimal code changes while gaining both speed and accuracy improvements. For teams already invested in R's statistical forecasting ecosystem, the tool offers comparable or faster performance than R implementations.
The project maintains active engagement with users through a community Slack channel. Development includes comprehensive documentation with end-to-end walkthroughs, tutorials covering anomaly detection, cross-validation, and multiple seasonalities, and published experiments demonstrating performance comparisons against alternative libraries. The codebase includes experiment directories with reproducible benchmarks showing relative performance across different model types and scales.
