Module feature_engineering 

Feature Engineering Pipelines for Stream Processing

This module provides a comprehensive feature engineering framework for real-time stream processing, enabling automatic feature extraction, transformation, and selection for machine learning workflows.

Features

• Automatic feature extraction from streaming events
• Real-time feature transformations (scaling, encoding, binning)
• Time-based features (rolling windows, lag features, rate of change)
• Categorical encoding (one-hot, label, target encoding)
• Feature selection and dimensionality reduction
• Feature store integration for reusability
• Pipeline composition with visual DAG representation

Example Usage

use oxirs_stream::feature_engineering::{FeaturePipeline, FeatureTransform};

let mut pipeline = FeaturePipeline::new();
pipeline
    .add_transform(FeatureTransform::StandardScaler)
    .add_transform(FeatureTransform::RollingMean { window: 10 })
    .add_transform(FeatureTransform::OneHotEncoder { columns: vec!["category".into()] });

let features = pipeline.transform(&event)?;

Structs

Feature

Feature definition

FeatureExtractionConfig

Feature extraction configuration

FeatureMetadata

Feature metadata

FeaturePipeline

Feature engineering pipeline

FeatureSet

Feature set (collection of features)

FeatureStore

Feature store for reusable features

PipelineStats

Pipeline statistics

Enums

FeatureTransform

Feature transformation types

FeatureValue

Feature data type

ImputationStrategy

Imputation strategy for missing values