Module outlier_transformation 

Outlier transformation methods for handling extreme values

This module provides various transformation methods specifically designed to handle outliers in data while preserving the overall structure and relationships. Unlike outlier detection which identifies outliers, these methods transform them to reduce their impact on downstream analysis.

Features

• Log Transformation: Reduces impact of large outliers through logarithmic scaling
• Square Root Transformation: Mild transformation for positive outliers
• Box-Cox Transformation: Data-driven power transformation for normalization
• Quantile Transformation: Maps to uniform or normal distribution
• Robust Scaling: Scaling resistant to outliers using median and IQR
• Outlier Interpolation: Replace outliers with interpolated values
• Outlier Smoothing: Smooth outliers using neighboring values
• Trimmed Transformation: Apply transformations after trimming extreme percentiles

Structs

FeatureTransformationParams

Transformation parameters for a single feature

GlobalTransformationParams

Global transformation parameters

OutlierTransformationConfig

Configuration for outlier transformation

OutlierTransformer

Outlier transformer for handling extreme values through transformation

TransformationParameters

Parameters learned during fitting for transformations

Enums

OutlierTransformationMethod

Available outlier transformation methods