redpandas-ml 0.3.2

Enterprise-grade ML data-preparation pipeline with integrated visual health reporting

redpandas-prep

Enterprise-grade ML data-preparation pipeline with integrated visual health reporting.

Installation

# Core only
pip install redpandas-prep

# With visualization support (recommended)
pip install "redpandas-prep[viz]"

# With all optional dependencies (seaborn, scipy, scikit-learn)
pip install "redpandas-prep[all]"

Note: The package name on PyPI is redpandas-prep, but you import it as redpandas using the alias rp:

import redpandas as rp

Quick Start (Chain Usage)

import redpandas as rp

# Full automated pipeline
X, y = (
    rp.Prep("messy_data.csv")
      .clean()
      .transform()
      .plot_report(target_column="label", save_path="report.html")
      .split_target("label")
)

Configuration (PrepConfig)

You can fully customize the pipeline behavior using the new PrepConfig class:

from redpandas.config import PrepConfig

config = PrepConfig(
    missing_threshold=0.80,
    outlier_method="iqr",
    scaling_method="robust",
    categorical_imputation="mode",
    auto_encode=True
)

prep = rp.Prep("messy_data.csv", config=config)

Diagnostics & Health Scoring

redpandas automatically diagnoses datasets upon initialization:

print(prep.diagnostics["health_score"])
print(prep.diagnostics["issues"])
print(prep.diagnostics["recommendations"])

Data Cleaning (.clean())

Step	Action	Config Reference
1	Remove exact duplicate rows	remove_duplicates
2	Drop empty/sparse columns	missing_threshold
3	Drop zero-variance columns	remove_constant_columns
4	Remove statistical outliers	outlier_method

Data Transformation (.transform())

Step	Action	Config Reference
1	Impute numeric columns	numerical_imputation
2	Impute categorical columns	categorical_imputation
3	Scale numeric columns	scaling_method
4	One-hot encode categories	auto_encode

Pipeline Persistence

Save and load the entire pipeline state:

prep.save_pipeline("pipeline.pkl")
loaded_prep = rp.Prep.load_pipeline("pipeline.pkl")

Scikit-Learn Integration

Export the exact configuration of redpandas to a scikit-learn ColumnTransformer:

sklearn_pipeline = prep.to_sklearn_pipeline()
# Use directly in GridSearchCV or other scikit-learn tools

Visualization

The .plot_report() method generates a single self-contained HTML report with zero external dependencies. The report contains:

• Dataset Overview & Data Health Score
• Memory Usage & Recommendations
• Missing Value Analysis
• Feature Distributions
• Outliers (Box-Whisker)
• Correlation Matrix
• Class Balance

Publishing a New Version

To publish to PyPI using GitHub Actions:

1. Update version in pyproject.toml
2. Create and push a new tag:

   git tag v0.3.0
   git push origin v0.3.0
   

3. GitHub Actions handles the build and trusted PyPI publishing automatically.