Polars Loess

This is a loess local regression (locally estimated scatterplot smoothing) implementation in Rust for Polars.

This is an early release. There is room for improvement in terms of performance, memory efficiency and feature set. I value feedback from the community to see where to go next.

Installation

pip install polars-loess

API

loess(
    'x-column',     # Name of the existing x-values
    'y-column',     # Name of the existing y-values
    'new-x-column', # Name of the new x values. y-values are interpolated for these x-values using loess
    
    # Optional float to specify the fraction of the data used in each local regression.
    # Exactly one of frac or points must be specified.
    frac=None,
    
    # Optional integer to specify the number of points used in each local regression.
    # Exactly one of frac or points must be specified.
    points=None,
    
    # Optional integer to specify the degree of the polynomial used in each local regression.
    # Default is 1.
    degree=None,
)

Example

import polars as pl
from polars_loess import loess

df = pl.DataFrame({
    'time': [
        0.5578196, 2.0217271, 2.5773252, 3.4140288, 4.3014084,
        4.7448394, 5.1073781, 6.5411662, 6.7216176, 7.2600583,
        8.1335874, 9.1224379, 11.9296663, 12.3797674, 13.2728619,
        14.2767453, 15.3731026, 15.6476637, 18.5605355, 18.5866354
    ],
    'price': [
        18.63654, 103.49646, 150.35391, 190.51031, 208.70115,
        213.71135, 228.49353, 233.55387, 234.55054, 223.89225,
        227.68339, 223.91982, 168.01999, 164.95750, 152.61107,
        160.78742, 168.55567, 152.42658, 221.70702, 222.69040,
    ],
})
result = df.with_columns(loess = loess('time', 'price', 'time', frac=0.5))
print(result)

Another example can be found in run.py. The result looks like.

Acknowledgements

• The loess implementation is based on https://github.com/joaofig/loess-rs
• The Python bindings are based on https://marcogorelli.github.io/polars-plugins-tutorial/