Home / Market & alternative data
functime
Time-series machine learning at scale with Polars
What it is
A Python library for global forecasting and time-series feature extraction on large panel datasets. It provides preprocessing, cross-validation splitters, forecast metrics such as MASE and SMAPE, over 100 feature extractors, exogenous feature support, backtesting, and automated hyperparameter tuning, all built as lazy Polars transforms for parallel processing across many time series. Quant teams and data scientists working with large numbers of related series are the main audience. The project is distributed under Apache-2.0, and its LLM-powered forecast analyst requires installing optional extras.
At a glance
Worth watchingOur rating, based on popularity, maintenance and how ready it is for real use.
| Best for | Professional quants |
|---|---|
| Used for | Data analysis, Strategy research |
| Markets | Multi-market |
| Stack | Python (Polars) |
| Learning curve | Moderate learning curve |
| Practical value | Medium practical value |
| Cost | Free and open source |
| Hardware | No GPU needed |
| Maintenance | Last commit 5 months ago |
GitHub stars, last 30 days
Daily snapshots since 2026-09-12 (up to 30 days): +4 over the period, now 1,188. Gaps mean no snapshot was taken that day.
Time-series machine learning at scale. Built with Polars for embarrassingly parallel feature extraction and forecasts on panel data.
Similar tools
Jupyter notebooks companion to Machine Learning for Trading
Goldman Sachs Python toolkit for derivatives and risk
scikit-learn compatible Python library for portfolio optimization
CVXPY-based portfolio optimization and strategic asset allocation
Python financial econometrics with ARCH, GARCH and volatility models
C++ financial research terminal with embedded Python analytics