polarsmith 1.0.0

Automated feature engineering in one function, 100% Polars-native

Installation

pip install polarsmith
# atau dengan uv:
uv add polarsmith

Quick Start

import polars as pl
from polarsmith import forge

df = pl.DataFrame({
    "age": [25, 30, 35, 40],
    "income": [50000, 60000, 70000, 80000],
    "churn": [0, 1, 0, 1]
})

df_out = forge(
    df,
    target="churn",
    strategy="smart",          # auto-detect fitur yang relevan
    config={
        "binning":      {"max_bins": 10},
        "interactions": ["age*income"],
        "encoding":     {"method": "james_stein"},
    }
)

Feature Matrix

Fitur	Status
Auto Binning	✅ v0.1
Cyclical Features	✅ v0.1
Interaction Terms	✅ v0.1
James-Stein Enc.	✅ v0.1
WoE Encoding	✅ v0.1
Smart Detection	✅ v0.1
Pandas Compat	✅ v0.1

Performance

Benchmark pada MacBook Air M-series, Python 3.12:

Dataset Size	Polarsmith	Pandas equivalent	Speedup
10,000 rows	0.0xs	0.0xs	Xx
100,000 rows	0.0xs	0.0xs	Xx
1,000,000 rows	0.0xs	0.0xs	Xx

Polarsmith menggunakan Polars 1.x dengan lazy evaluation dan SIMD.