polars-timeseries 0.3.0

Polars Time Series Extension

Polars Time Series Extension

Welcome to the documentation for the Polars Time Series Extension.

---

Documentation: https://drumtorben.github.io/polars-ts

Source Code: https://github.com/drumtorben/polars-ts

---

📖 Overview

The Polars Time Series Extention offers a wide range of metrics, feature extractors, and various tools for time series forecasting.

Installation

pip install polars-timeseries

How to use

The polars-ts plugin is available under the namespace pts. See the following example where we compute the Kaboudan metric:

import polars as pl
from statsforecast import StatsForecast
from statsforecast.models import AutoETS, OptimizedTheta

import polars_ts as pts  # noqa

# Create sample dataframe with columns `unique_id`, `ds`, and `y`.
df = (
    pl.scan_parquet("https://datasets-nixtla.s3.amazonaws.com/m4-hourly.parquet")
    .filter(pl.col("unique_id").is_in(["H1", "H2", "H3"]))
    .collect()
)

# Define models
season_length = 24
models = [
    OptimizedTheta(season_length=season_length, decomposition_type="additive"),
    AutoETS(season_length=season_length),
]
sf = StatsForecast(models=models, freq=1, n_jobs=-1)

# Compute the Kaboudan metric in the `pts` namespace
res = df.pts.kaboudan(sf, block_size=200, backtesting_start=0.5, n_folds=10)