qpx-tabular 0.1.2

A powerful, production-ready tabular data preprocessing and visualization library.

QPX Tabular

QPX Tabular is a powerful, production-ready tabular data preprocessing and visualization library designed to accelerate data science workflows. It turns raw, messy pandas DataFrames into machine-learning ready datasets with a single line of code.

Features

• Automated Preprocessing (auto_preprocess): Automatically handles missing values, drops constants, drops high-cardinality nominals, encodes categoricals intelligently, and downcasts memory.
• Fail-Loud Architecture: Built for production. Instead of failing silently, QPX immediately alerts you (KeyError, ValueError) if you provide invalid data configurations.
• Comprehensive Data Health Diagnostics: Get 360-degree views of your dataset's health via dataset_health and statistical_snapshot.
• Beautiful Visualizations: One-line correlation heatmaps, distribution plots, and hierarchical feature clustering matrices.

Installation

To install qpx-tabular via PyPI (once published) or from source, you can simply clone this repository and install it locally using pip:

git clone https://github.com/punitxdev/QPX.git
cd QPX
pip install -e .

Dependencies

• pandas
• numpy
• matplotlib
• seaborn
• scipy

Quickstart

Clean an entire dataset with one function:

import pandas as pd
from qpx.tabular import preprocessing

# Load your raw data
df = pd.read_csv("my_messy_data.csv")

# Clean, encode, impute, and downcast in one go!
clean_df, report = preprocessing.auto_preprocess(
    df,
    max_onehot=10, 
    return_report=True
)

print(report)

Generate a deep-dive correlation map:

from qpx.tabular import visuals

visuals.corr_map(clean_df, target="my_target_column")

Documentation

The complete API reference and user guide is hosted online at: https://punitxdev.github.io/QPX/

If you want to build the documentation locally for development:

pip install -e .[dev]
mkdocs serve

To publish the documentation to GitHub Pages, simply run:

mkdocs gh-deploy

License

This project is licensed under the MIT License - see the LICENSE file for details.

Made with love by Punit