skpipeline-forge 1.0.0

Composable sklearn-compatible pipeline builder for feature engineering, validation, resampling, and monitoring.

skpipeline-forge

skpipeline-forge is a composable, sklearn-compatible pipeline builder for tabular ML workflows.

It helps you define reusable transformation stacks as objects, then apply them consistently across:

• training
• scoring
• optional imbalance handling
• monitoring/reporting hooks

Install

pip install skpipeline-forge

Quickstart

from pipeline_forge import FeaturePipelineBuilder

builder = FeaturePipelineBuilder(target_col="target", random_seed=42)

(
    builder
    .validate_schema(required_cols=["city", "hour", "age", "balance"], strict=False)
    .group_rare_categories(categorical_cols=["city"], min_freq=0.02)
    .target_encode(categorical_cols=["city"], drop_original=True)
    .encode_cyclical_time(period_map={"hour": 24})
    .add_interaction_features(numeric_cols=["age", "balance"])
    .scaler("robust")
    .random_oversampling(sampling_strategy=1.0)
)

builder.build_pipeline()

What You Get

• Fluent API for chaining transformations
• sklearn/imblearn-compatible pipelines
• Feature engineering + resampling in one object
• Extra pipeline layers for validation, selection, postprocessing, monitoring, explainability
• Built-in common transformers for tabular data

Documentation

• Overview
• Pipeline Layers
• Transformation Catalog
• Examples
• Tutorial Notebook

Package Notes

• Python: >=3.10,<4.0
• Main runtime deps: scikit-learn, imbalanced-learn, pandas, numpy, mlflow, prophet
• Tests: pytest

Development

poetry install --with dev
poetry run pytest -q
poetry build