Tab-right

Tab-right is a Python package for easy analysis of tabular data for inference models (ML and non-ML), focusing on model-agnostic diagnostics using predictions.

Getting Started

• Introduction

Guides & Examples

• Architecture
• Analysis Guides

API Reference

• API Reference
  • Segmentation
  • Drift Detection
  • Plotting Functions
  • Base Architecture & Protocols
  • Task Detection

Quickstart

Install tab-right using pip:

pip install tab-right

Basic Example:

import pandas as pd
import numpy as np
from tab_right.drift.drift_calculator import DriftCalculator
from tab_right.plotting.drift_plotter import DriftPlotter

# Create sample datasets
df_reference = pd.DataFrame({
    'numeric_feature': np.random.normal(0, 1, 100),
    'categorical_feature': np.random.choice(['A', 'B', 'C'], 100)
})

df_current = pd.DataFrame({
    'numeric_feature': np.random.normal(0.5, 1.2, 100),
    'categorical_feature': np.random.choice(['A', 'B', 'C'], 100, p=[0.6, 0.3, 0.1])
})

# Calculate and visualize drift
drift_calc = DriftCalculator(df_reference, df_current)
drift_plotter = DriftPlotter(drift_calc)

# Get drift metrics and visualize results
drift_results = drift_calc()
drift_plot = drift_plotter.plot_multiple()

Advanced Usage

• Feature types: For continuous features, set is_categorical=False and optionally adjust the bins parameter.

• Multi-class outputs: For multi-class or probabilistic outputs, pass a list of label columns and use probability mode.

• Custom metrics: Use any metric function compatible with scikit-learn (e.g., accuracy_score, r2_score).

• Drift thresholds: Set custom thresholds for different severity levels with the thresholds parameter.

• Visualization options: Customize plots with matplotlib parameters and custom color schemes.

Contributing

See the CONTRIBUTING.md file for guidelines.

License

MIT License